This paper presents a deep neural solver to automatically solve math word problems. In contrast to previous statistical learning approaches, we directly translate math word problems to equation templates using a recurrent neural network (RNN) model, without sophisticated feature engineering. We further design a hybrid model that combines the RNN model and a similarity-based retrieval model to achieve additional performance improvement. Experiments conducted on a large dataset show that the RNN model and the hybrid model significantly outperform stateof-the-art statistical learning methods for math word problem solving.

Bioinspired re-entrant structures have been proved to be effective in achieving liquid superrepellence (including anti-penetration, anti-adhesion, and anti-spreading). However, except for a few reports relying on isotropic etching of silicon wafers, most fluorination-dependent surfaces are still unable to repel liquids with extreme low surface energy (i.e., γ < 15 mN m ), especially those fluorinated solvents. Herein, triply re-entrant structures, possessing superrepellence to water (with surface tension γ of 72.8 mN m ) and various organic liquids (γ = 12.0-27.1 mN m ), are fabricated via two-photon polymerization based 3D printing technology. Such structures can be constructed both on rigid and flexible substrates, and the liquid superrepellent properties can be kept even after oxygen plasma treatment. Based on the prepared triply re-entrant structures, micro open capillaries are constructed on them to realize directional liquid spreading, which may be applied in microfluidic platforms and lab-on-a-chip applications. The fabricated arrays can also find potential applications in electronic devices, gas sensors, microchemical/physical reactors, high-throughput biological sensors, and optical displays.

This paper presents a semantic parsing and reasoning approach to automatically solving math word problems. A new meaning representation language is designed to bridge natural language text and math expressions. A CFG parser is implemented based on 9,600 semi-automatically created grammar rules. We conduct experiments on a test set of over 1,500 number word problems (i.e., verbally expressed number problems) and yield 95.4% precision and 60.2% recall.

Multiresponsive elastic poly(methyl methacrylate-butyl acrylate) (P(MMA-BA)) copolymer nanoparticles with controlled sizes are fabricated through a onestep method, which further serve as building blocks for the construction of multiresponsive films via self-assembly. Taking advantage of the relatively low glass transition temperature and the core-shell structure of the copoly mer nanoparticles, they possess the capacity to partially deform and fuse at room temperature under dry status, eventually resulting in the enhancement of the mechanical properties as well as the control of optical properties in the assembled ordered structures. The generated elastic films not only can control the concealment or exhibition of the designed color information, but also can rapidly respond to external stimuli such as the solvent, pH, and tensile force in a reversible fashion. These functional elastic copolymer nanoparticles have potential applications in dynamic color display, optical sensing, and anticounterfeiting.

Sequence-to-sequence (SEQ2SEQ) models have been successfully applied to automatic math word problem solving. Despite its simplicity, a drawback still remains: a math word problem can be correctly solved by more than one equations. This non-deterministic transduction harms the performance of maximum likelihood estimation. In this paper, by considering the uniqueness of expression tree, we propose an equation normalization method to normalize the duplicated equations. Moreover, we analyze the performance of three popular SEQ2SEQ models on the math word problem solving. We find that each model has its own specialty in solving problems, consequently an ensemble model is then proposed to combine their advantages. Experiments on dataset Math23K show that the ensemble model with equation normalization significantly outperforms the previous state-of-the-art methods.