Cloze-style reading comprehension is a representative problem in mining relationship between document and query. In this paper, we present a simple but novel model called attention-over-attention reader for better solving cloze-style reading comprehension task. The proposed model aims to place another attention mechanism over the document-level attention and induces "attended attention" for final answer predictions. One advantage of our model is that it is simpler than related works while giving excellent performance. In addition to the primary model, we also propose an N-best re-ranking strategy to double check the validity of the candidates and further improve the performance. Experimental results show that the proposed methods significantly outperform various state-ofthe-art systems by a large margin in public datasets, such as CNN and Children's Book Test.

Developing effective catalysts based on earth abundant elements is critical for CO 2 electroreduction. However, simultaneously achieving a high Faradaic efficiency (FE) and high current density of CO (j CO) remains a challenge. Herein, we prepare a Mn single-atom catalyst (SAC) with a Mn-N 3 site embedded in graphitic carbon nitride. The prepared catalyst exhibits a 98.8% CO FE with a j CO of 14.0 mA cm −2 at a low overpotential of 0.44 V in aqueous electrolyte, outperforming all reported Mn SACs. Moreover, a higher j CO of 29.7 mA cm −2 is obtained in an ionic liquid electrolyte at 0.62 V overpotential. In situ X-ray absorption spectra and density functional theory calculations demonstrate that the remarkable performance of the catalyst is attributed to the Mn-N 3 site, which facilitates the formation of the key intermediate COOH * through a lowered free energy barrier.

Large numbers of catalysts have been developed for the electrochemical reduction of CO to value-added liquid fuels. However, it remains a challenge to maintain a high current efficiency in a wide negative potential range for achieving a high production rate of the target products. Herein, we report a 2D/0D composite catalyst composed of bismuth oxide nanosheets and nitrogen-doped graphene quantum dots (Bi O -NGQDs) for highly efficient electrochemical reduction of CO to formate. Bi O -NGQDs demonstrates a nearly 100 % formate Faraday efficiency (FE) at a moderate overpotential of 0.7 V with a good stability. Strikingly, Bi O -NGQDs exhibit a high activity (average formate FE of 95.6 %) from -0.9 V to -1.2 V vs. RHE. Additionally, DFT calculations reveal that the origin of enhanced activity in this wide negative potential range can be attributed to the increased adsorption energy of CO (ads) and OCHO* intermediate after combination with NGQDs.