scholarly journals He Thinks He Knows Better than the Doctors: BERT for Event Factuality Fails on Pragmatics

2021 ◽  
Vol 9 ◽  
pp. 1081-1097
Author(s):  
Nanjiang Jiang ◽  
Marie-Catherine de Marneffe

Abstract We investigate how well BERT performs on predicting factuality in several existing English datasets, encompassing various linguistic constructions. Although BERT obtains a strong performance on most datasets, it does so by exploiting common surface patterns that correlate with certain factuality labels, and it fails on instances where pragmatic reasoning is necessary. Contrary to what the high performance suggests, we are still far from having a robust system for factuality prediction.

Author(s):  
R. Levi-Setti ◽  
J. M. Chabala ◽  
R. Espinosa ◽  
M. M. Le Beau

We have shown previously that isotope-labelled nucleotides in human metaphase chromosomes can be detected and mapped by imaging secondary ion mass spectrometry (SIMS), using the University of Chicago high resolution scanning ion microprobe (UC SIM). These early studies, conducted with BrdU- and 14C-thymidine-labelled chromosomes via detection of the Br and 28CN- (14C14N-> labelcarrying signals, provided some evidence for the condensation of the label into banding patterns along the chromatids (SIMS bands) reminiscent of the well known Q- and G-bands obtained by conventional staining methods for optical microscopy. The potential of this technique has been greatly enhanced by the recent upgrade of the UC SIM, now coupled to a high performance magnetic sector mass spectrometer in lieu of the previous RF quadrupole mass filter. The high transmission of the new spectrometer improves the SIMS analytical sensitivity of the microprobe better than a hundredfold, overcoming most of the previous imaging limitations resulting from low count statistics.


SLEEP ◽  
2020 ◽  
Author(s):  
Evan D Chinoy ◽  
Joseph A Cuellar ◽  
Kirbie E Huwa ◽  
Jason T Jameson ◽  
Catherine H Watson ◽  
...  

Abstract Study Objectives Consumer sleep-tracking devices are widely used and becoming more technologically advanced, creating strong interest from researchers and clinicians for their possible use as alternatives to standard actigraphy. We therefore tested the performance of many of the latest consumer sleep-tracking devices, alongside actigraphy, versus the gold-standard sleep assessment technique, polysomnography (PSG). Methods In total, 34 healthy young adults (22 women; 28.1 ± 3.9 years, mean ± SD) were tested on three consecutive nights (including a disrupted sleep condition) in a sleep laboratory with PSG, along with actigraphy (Philips Respironics Actiwatch 2) and a subset of consumer sleep-tracking devices. Altogether, four wearable (Fatigue Science Readiband, Fitbit Alta HR, Garmin Fenix 5S, Garmin Vivosmart 3) and three non-wearable (EarlySense Live, ResMed S+, SleepScore Max) devices were tested. Sleep/wake summary and epoch-by-epoch agreement measures were compared with PSG. Results Most devices (Fatigue Science Readiband, Fitbit Alta HR, EarlySense Live, ResMed S+, SleepScore Max) performed as well as or better than actigraphy on sleep/wake performance measures, while the Garmin devices performed worse. Overall, epoch-by-epoch sensitivity was high (all ≥0.93), specificity was low-to-medium (0.18-0.54), sleep stage comparisons were mixed, and devices tended to perform worse on nights with poorer/disrupted sleep. Conclusions Consumer sleep-tracking devices exhibited high performance in detecting sleep, and most performed equivalent to (or better than) actigraphy in detecting wake. Device sleep stage assessments were inconsistent. Findings indicate that many newer sleep-tracking devices demonstrate promising performance for tracking sleep and wake. Devices should be tested in different populations and settings to further examine their wider validity and utility.


2020 ◽  
Vol 34 (07) ◽  
pp. 10607-10614 ◽  
Author(s):  
Xianhang Cheng ◽  
Zhenzhong Chen

Learning to synthesize non-existing frames from the original consecutive video frames is a challenging task. Recent kernel-based interpolation methods predict pixels with a single convolution process to replace the dependency of optical flow. However, when scene motion is larger than the pre-defined kernel size, these methods yield poor results even though they take thousands of neighboring pixels into account. To solve this problem in this paper, we propose to use deformable separable convolution (DSepConv) to adaptively estimate kernels, offsets and masks to allow the network to obtain information with much fewer but more relevant pixels. In addition, we show that the kernel-based methods and conventional flow-based methods are specific instances of the proposed DSepConv. Experimental results demonstrate that our method significantly outperforms the other kernel-based interpolation methods and shows strong performance on par or even better than the state-of-the-art algorithms both qualitatively and quantitatively.


2019 ◽  
Vol 485 (3) ◽  
pp. 3370-3377 ◽  
Author(s):  
Lehman H Garrison ◽  
Daniel J Eisenstein ◽  
Philip A Pinto

Abstract We present a high-fidelity realization of the cosmological N-body simulation from the Schneider et al. code comparison project. The simulation was performed with our AbacusN-body code, which offers high-force accuracy, high performance, and minimal particle integration errors. The simulation consists of 20483 particles in a $500\ h^{-1}\, \mathrm{Mpc}$ box for a particle mass of $1.2\times 10^9\ h^{-1}\, \mathrm{M}_\odot$ with $10\ h^{-1}\, \mathrm{kpc}$ spline softening. Abacus executed 1052 global time-steps to z = 0 in 107 h on one dual-Xeon, dual-GPU node, for a mean rate of 23 million particles per second per step. We find Abacus is in good agreement with Ramses and Pkdgrav3 and less so with Gadget3. We validate our choice of time-step by halving the step size and find sub-percent differences in the power spectrum and 2PCF at nearly all measured scales, with ${\lt }0.3{{\ \rm per\ cent}}$ errors at $k\lt 10\ \mathrm{Mpc}^{-1}\, h$. On large scales, Abacus reproduces linear theory better than 0.01 per cent. Simulation snapshots are available at http://nbody.rc.fas.harvard.edu/public/S2016.


1998 ◽  
Vol 514 ◽  
Author(s):  
D. Edelstein

ABSTRACTRecently IBM announced the first implementation of full copper ULSI wiring in a CMOS technology, to be manufactured on its high-performance 0.22 um CMOS products this year. Features of this technology will be presented, as well as functional verification on CMOS chips. To reach this level, extensive yield, reliability, and stress testing had to be done on test and product-like chips, including those packaged into product modules. Data will be presented fom all aspects of this testing, ranging from experiments designed to promote Cu contamination of the MOS devices, to temperature/humidity/bias stressing of assembled functional modules. The results in all areas are shown to be equal to or better than standards set by our current AI(Cu)/Wstud technology. This demonstrates that the potential problems associated with copper wiring that have long been discussed can be overcome.


Author(s):  
P. Laurent ◽  
F. Acero ◽  
V. Beckmann ◽  
S. Brandt ◽  
F. Cangemi ◽  
...  

AbstractBased upon dual focusing techniques, the Polarimetric High-Energy Modular Telescope Observatory (PHEMTO) is designed to have performance several orders of magnitude better than the present hard X-ray instruments, in the 1–600 keV energy range. This, together with its angular resolution of around one arcsecond, and its sensitive polarimetry measurement capability, will give PHEMTO the improvements in scientific performance needed for a mission in the 2050 era in order to study AGN, galactic black holes, neutrons stars, and supernovae. In addition, its high performance will enable the study of the non-thermal processes in galaxy clusters with an unprecedented accuracy.


2021 ◽  
Vol 0 (0) ◽  
Author(s):  
Hang Li ◽  
Jiamin Liu ◽  
Zhanzhong Wang ◽  
Xiaodong Liu ◽  
Xichun Yan ◽  
...  

Abstract With chili and liquid beef tallow as the main raw materials, the processing conditions of chili flavor beef tallow were explored. Gas chromatograpy-ion mobility spectrometry (GC-IMS) was used to determine the volatile compounds in chili flavor beef tallow. The capsaicin and dihydrocapsaicin in chili flavor beef tallow were determined by high performance liquid chromatography (HPLC). The optimum technological conditions were determined, and the index of chromatic aberration, cholesterol was also determined. Based on GC-IMS analysis, 102 kinds of volatile compounds were detected, and the sample III (the ratio of solid–liquid was 1:5, the frying temperature was 120 °C, and the frying time was 15 min) performed better than other samples. The preparation of chili beef tallow improves its antioxidant activity and makes its aroma more intense and more in line with the taste of Chinese people, which provides a theoretical and practical basis for the development of spice beef tallow in the future.


2021 ◽  
Author(s):  
Lucas Bragança ◽  
Jeronimo Penha ◽  
Michael Canesche ◽  
Dener Ribeiro ◽  
José Augusto M. Nacif ◽  
...  

FPGAs are suitable to speed up gene regulatory network (GRN) algorithms with high throughput and energy efficiency. In addition, virtualizing FPGA using hardware generators and cloud resources increases the computing ability to achieve on-demand accelerations across multiple users. Recently, Amazon AWS provides high-performance Cloud's FPGAs. This work proposes an open source accelerator generator for Boolean gene regulatory networks. The generator automatically creates all hardware and software pieces from a high-level GRN description. We evaluate the accelerator performance and cost for CPU, GPU, and Cloud FPGA implementations by considering six GRN models proposed in the literature. As a result, the FPGA accelerator is at least 12x faster than the best GPU accelerator. Furthermore, the FPGA reaches the best performance per dollar in cloud services, at least 5x better than the best GPU accelerator.


Author(s):  
Bryan J Starkey ◽  
Ian D R Fry

A new commercial enzymic kit for urinary oxalate determination has been adapted for use on a centrifugal analyser. It has been evaluated and compared with an established high performance liquid chromatographic (HPLC) procedure developed in our laboratory. Mean recovery of oxalate from urine samples augmented with oxalic acid exceeded 97% by both methods. The precision of the HPLC method was superior to that of the enzymic kit but both methods gave between batch precision values better than CV 12% at low (less than 100μmol/L) oxalate concentrations and better than CV 7% at higher concentrations (greater than 270μmol/L) Urinary oxalate values obtained with the new enzymic procedure correlated more closely with HPLC values ( r = 0·84) than did values previously obtained using the forerunner of the kit ( r = 0·62) which was known to be susceptible to ascorbate interference. No significant interference from ascorbic acid or from high urinary calcium concentrations could be demonstrated using either the improved kit or the HPLC procedure. Its easy adaptation to automated analysers available in most laboratories, coupled to its acceptable analytical performance render the enzymic kit a reasonable alternative to HPLC or other more complex procedures for urinary oxalate analysis.


1970 ◽  
Vol 46 (4) ◽  
pp. 325-328
Author(s):  
A. H. Teich ◽  
M. J. Holst

Scots pine provenances from Russia were grown in Ontario at Longlac, Dorset and Chalk River, and in Saskatchewan at Prince Albert and Indian Head. After eight growing seasons in the field all provenances survived well. The three tallest, Orel, Woronesh and Kiev provenances, grew considerably taller than the others, and at Prince Albert, where native jack pine was also grown, the growth and survival of these provenances was also better than that of jack pine. If further studies show that this high performance persists and that stem form is commercially acceptable, seedlings from these three provenances may be useful for reforestation in those parts of Canada with a continental climate.


Sign in / Sign up

Export Citation Format

Share Document