In multi-turn dialogue, utterances do not always take the full form of sentences. These incomplete utterances will greatly reduce the performance of open-domain dialogue systems. Restoring more incomplete utterances from context could potentially help the systems generate more relevant responses. To facilitate the study of incomplete utterance restoration for open-domain dialogue systems, a large-scale multi-turn dataset Restoration-200K 1 is collected and manually labeled with the explicit relation between an utterance and its context. We also propose a "pick-andcombine" model to restore the incomplete utterance from its context. Experimental results demonstrate that the annotated dataset and the proposed approach significantly boost the response quality of both single-turn and multiturn dialogue systems.

Recurrent Neural Networks (RNNs) and their variants, such as Long-Short Term Memory (LSTM) networks, and Gated Recurrent Unit (GRU) networks, have achieved promising performance in sequential data modeling. The hidden layers in RNNs can be regarded as the memory units, which are helpful in storing information in sequential contexts. However, when dealing with high dimensional input data, such as video and text, the input-to-hidden linear transformation in RNNs brings high memory usage and huge computational cost. This makes the training of RNNs very difficult. To address this challenge, we propose a novel compact LSTM model, named as TR-LSTM, by utilizing the low-rank tensor ring decomposition (TRD) to reformulate the input-to-hidden transformation. Compared with other tensor decomposition methods, TR-LSTM is more stable. In addition, TR-LSTM can complete an end-to-end training and also provide a fundamental building block for RNNs in handling large input data. Experiments on real-world action recognition datasets have demonstrated the promising performance of the proposed TR-LSTM compared with the tensortrain LSTM and other state-of-the-art competitors.

No abstract

Human-machine interfaces (HMIs) are widely studied to understand the human biomechanics and/or physiology and the interaction between humans and machines/robots. The conventional rigid or invasive HMIs that record/send information from/to human bodies have significant disadvantages in practice for longterm, portable, and comfortable usages. To better adapt to natural soft skins, soft HMIs have been designed to deform into arbitrary shapes, and their bendable, stretchable, compressible and twistable properties offer a huge potential in future personalized applications. This paper presents a survey on various soft HMIs in terms of design, sensing, stimulation as well as their applications. Specifically, tactile / motion / bio-potential sensors are categorized for recording various data from human bodies, while stimulators are discussed for information feedback and motion activation to human bodies. It is anticipated that soft HMIs will promote the interaction among humans, machines/robots and environment to achieve desired coexisting-cooperativecognitive function in a robot system, named as Tri-Co Robot, for the human-centered applications, such as rehabilitation, medical monitoring and human-robot cooperation.

Natural Language Sentence Matching (NLSM) has gained substantial attention from both academics and the industry, and rich public datasets contribute a lot to this process. However, biased datasets can also hurt the generalization performance of trained models and give untrustworthy evaluation results. For many NLSM datasets, the providers select some pairs of sentences into the datasets, and this sampling procedure can easily bring unintended pattern, i.e., selection bias. One example is the QuoraQP dataset, where some content-independent naïve features are unreasonably predictive. Such features are the reflection of the selection bias and termed as the "leakage features." In this paper, we investigate the problem of selection bias on six NLSM datasets and find that four out of them are significantly biased. We further propose a training and evaluation framework to alleviate the bias. Experimental results on QuoraQP suggest that the proposed framework can improve the generalization ability of trained models, and give more trustworthy evaluation results for real-world adoptions.

No abstract

Large, 3D curved electronics are a trend of the microelectronic industry due to their unique ability to conformally coexist with complex surfaces while retaining the electronic functions of 2D planar integrated circuit technologies. However, these curved electronics present great challenges to the fabrication processes. Here, we propose a reconfigurable, mask-free, conformal fabrication strategy with a robot-like system, called robotized ‘transfer-and-jet’ printing, to assemble diverse electronic devices on complex surfaces. This novel method is a ground-breaking advance with the unique capability to integrate rigid chips, flexible electronics, and conformal circuits on complex surfaces. Critically, each process, including transfer printing, inkjet printing, and plasma treating, are mask-free, digitalized, and programmable. The robotization techniques, including measurement, surface reconstruction and localization, and path programming, break through the fundamental constraints of 2D planar microfabrication in the context of geometric shape and size. The transfer printing begins with the laser lift-off of rigid chips or flexible electronics from donor substrates, which are then transferred onto a curved surface via a dexterous robotic palm. Then the robotic electrohydrodynamic printing directly writes submicrometer structures on the curved surface. Their permutation and combination allow versatile conformal microfabrication. Finally, robotized hybrid printing is utilized to successfully fabricate a conformal heater and antenna on a spherical surface and a flexible smart sensing skin on a winged model, where the curved circuit, flexible capacitive and piezoelectric sensor arrays, and rigid digital–analog conversion chips are assembled. Robotized hybrid printing is an innovative printing technology, enabling additive, noncontact and digital microfabrication for 3D curved electronics.

No abstract