Speaker diarization has been mainly developed based on the clustering of speaker embeddings. However, the clustering-based approach has two major problems; i.e., (i) it is not optimized to minimize diarization errors directly, and (ii) it cannot handle speaker overlaps correctly. To solve these problems, the End-to-End Neural Diarization (EEND), in which a bidirectional long short-term memory (BLSTM) network directly outputs speaker diarization results given a multi-talker recording, was recently proposed. In this study, we enhance EEND by introducing self-attention blocks instead of BLSTM blocks. In contrast to BLSTM, which is conditioned only on its previous and next hidden states, self-attention is directly conditioned on all the other frames, making it much suitable for dealing with the speaker diarization problem. We evaluated our proposed method on simulated mixtures, real telephone calls, and real dialogue recordings. The experimental results revealed that the self-attention was the key to achieving good performance and that our proposed method performed significantly better than the conventional BLSTM-based method. Our method was even better than that of the state-of-the-art x-vector clustering-based method. Finally, by visualizing the latent representation, we show that the self-attention can capture global speaker characteristics in addition to local speech activity dynamics. Our source code is available online at https://github.com/hitachi-speech/EEND. Index Termsspeaker diarization, neural network, end-to-end, self-attention arXiv:1909.06247v1 [eess.AS] 13 Sep 2019 SAD MFCC X-vector extraction PLDA scoring AHC SAD neural network X-vector neural network Same/Diff covariance matrices Diarization result (a) X-vector clustering-based method Log-Mel Joint speech activity detection of all speakers EEND neural network Diarization result (b) EEND method

No abstract

Precipitation-hardening high-entropy alloys (PH-HEAs) with good strength−ductility balances are a promising candidate for advanced structural applications. However, current HEAs emphasize near-equiatomic initial compositions, which limit the increase of intermetallic precipitates that are closely related to the alloy strength. Here we present a strategy to design ultrastrong HEAs with high-content nanoprecipitates by phase separation, which can generate a near-equiatomic matrix in situ while forming strengthening phases, producing a PH-HEA regardless of the initial atomic ratio. Accordingly, we develop a non-equiatomic alloy that utilizes spinodal decomposition to create a low-misfit coherent nanostructure combining a near-equiatomic disordered face-centered-cubic (FCC) matrix with high-content ductile Ni3Al-type ordered nanoprecipitates. We find that this spinodal order–disorder nanostructure contributes to a strength increase of ~1.5 GPa (>560%) relative to the HEA without precipitation, achieving one of the highest tensile strength (1.9 GPa) among all bulk HEAs reported previously while retaining good ductility (>9%).

The critical bottleneck of electrocatalytic CO 2 reduction reaction (CO 2 RR) lies in its low efficiency at high overpotential caused by competitive hydrogen evolution. It is challenging to develop an efficient catalyst achieving both high current density and high Faradaic efficiency (FE) for CO 2 RR. Herein, we synthesized fluorine-doped cagelike porous carbon (F-CPC) by purposely tailoring its structural properties. The optimized F-CPC possesses large surface area with moderate mesopore and abundant micropores as well as high electrical conductivity. When used as catalyst for CO 2 RR, F-CPC exhibits FE of 88.3% for CO at −1.0 V vs RHE with a current density of 37.5 mA•cm −2 . Experimental results and finite element simulations demonstrate that the excellent CO 2 RR performance of F-CPC at high overpotential should be attributed to its structure-enhanced electrocatalytic process stemming from its cagelike morphology.