Previous work on dialog act (DA) classification has investigated different methods, such as hidden Markov models, maximum entropy, conditional random fields, graphical models, and support vector machines. A few recent studies explored using deep learning neural networks for DA classification, however, it is not clear yet what is the best method for using dialog context or DA sequential information, and how much gain it brings. This paper proposes several ways of using context information for DA classification, all in the deep learning framework. The baseline system classifies each utterance using the convolutional neural networks (CNN). Our proposed methods include using hierarchical models (recurrent neural networks (RNN) or CNN) for DA sequence tagging where the bottom layer takes the sentence CNN representation as input, concatenating predictions from the previous utterances with the CNN vector for classification, and performing sequence decoding based on the predictions from the sentence CNN model. We conduct thorough experiments and comparisons on the Switchboard corpus, demonstrate that incorporating context information significantly improves DA classification, and show that we achieve new state-of-the-art performance for this task.

In the last two decades, advances in micro‐ and nanofabrication have enabled the development of a wide variety of active or “self‐propelling” microparticles, which convert energy from their environment into directed motion. While these autonomous entities have shown promise for efficient locomotion on the microscale, their practical utility remains unrealized due to their inability to perform multiple useful tasks on demand. From an engineering perspective, the active particle behavior can be encoded on an individual level by tailoring key design elements such as shape, polarizability, surface pattern, and bulk functionality. This feature article focusses on active particles powered by electric and magnetic fields, as these sources of energy allow the particles to: (1) move in several phenomenologically unique ways, (2) respond in a reliable manner to the field parameters, and (3) interact synergistically to enable multiple functions. It is hypothesized how future generations of such particles may remotely harvest and transduce energy to perform several useful tasks such as biosensing and delivering drugs. As a step toward realizing such particles, several new types of active particles are demonstrated. Finally, a perspective on the future directions of this emerging field is provided by discussing current challenges, potential applications as well as future opportunities.

Sequence-encoded assembly of patchy magnetic microcubes enables making self-reconfiguring colloidal origami and “microbots.”

Active matter, both synthetic and biological, demonstrates complex spatiotemporal self-organization and the emergence of collective behavior. A coherent rotational motion, the vortex phase, is of great interest because of its ability to orchestrate well-organized motion of self-propelled particles over large distances. However, its generation without geometrical confinement has been a challenge. Here, we show by experiments and computational modeling that concentrated magnetic rollers self-organize into multivortex states in an unconfined environment. We find that the neighboring vortices more likely occur with the opposite sense of rotation. Our studies provide insights into the mechanism for the emergence of coherent collective motion on the macroscale from the coupling between microscale rotation and translation of individual active elements. These results may stimulate design strategies for self-assembled dynamic materials and microrobotics.