In recent years, large neural networks for natural language generation (NLG) have made leaps and bounds in their ability to generate fluent text. However, the tasks of evaluating quality differences between NLG systems and understanding how humans perceive the generated text remain both crucial and difficult. In this system demonstration, we present Real or Fake Text (RoFT), a website that tackles both of these challenges by inviting users to try their hand at detecting machine-generated text in a variety of domains. We introduce a novel evaluation task based on detecting the boundary at which a text passage that starts off humanwritten transitions to being machine-generated. We show preliminary results of using RoFT to evaluate detection of machine-generated news articles.

Mechanospray ionization (MoSI) is a technique that produces ions directly from solution-like electrospray ionization (ESI) but without the need of a high voltage. In MoSI, mechanical vibrations aerosolize solution phase analytes, whereby the resulting microdroplets can be directed into the inlet orifice of a mass spectrometer. In this work, MoSI is applied to biomolecules up to 80 kDa in mass in both denatured and native conditions as well as polymers up to 12 kDa in mass. The various MoSI devices used in these analyses were all comprised of a piezoelectric annulus attached to a central metallic disk containing an array of 4 to 7 μm diameter holes. The devices vibrated in the 100–170 kHz range to generate a beam of microdroplets that ultimately resulted in ion formation. A linear quadrupole ion trap (LIT) and orbitrap mass spectrometer were used in the analysis to investigate higher mass proteins at both native (folded) and denatured (unfolded) conditions. MoSI native mass spectra of proteins acquired on the orbitrap and LIT instrument demonstrated that proteins could remain intact and in a folded state. In the case of native MS of holomyoglobin, the intact folded protein remained mostly bound noncovalently to the heme group, and typically, the spectra showed reduced loss of the heme group by MoSI as compared to ESI. In both non-native and native protein analyses examples, broader often multimodal distributions to lower charge states were observed. When using the LIT instrument, a significant increase in the relative abundance of dimers was observed by MoSI as compared to ESI. The softness of the MoSI technique was evidenced by the lack of fragmentation, the formation of dimers as also noted by others (J. Mass Spectrom.2016424429) and under native conditions, the retention of proteins in one or more presumed folded structures and for holomyoglobin the high retention of the heme group. When analyzing polyethylene glycol (PEG) and polypropylene glycol (PPG), MoSI also generated a broader distribution to lower charge states than ESI. By using the improved separation of peaks at lower charge states and all the charge states available, MoSI data should provide an improved ionization method to obtain more accurate mass and dispersity values for some polymers.

Internet of Things(IoT) devices, mobile phones, and robotic systems are often denied the power of deep learning algorithms due to their limited computing power. However, to provide time critical services such as emergency response, home assistance, surveillance, etc, these devices often need real time analysis of their camera data. This paper strives to offer a viable approach to integrate high-performance deep learning based computer vision algorithms with low-resource and low-power devices by leveraging the computing power of the cloud. By offloading the computation work to the cloud, no dedicated hardware is needed to enable deep neural networks on existing low computing power devices. A Raspberry Pi based robot, Cloud Chaser, is built to demonstrate the power of using cloud computing to perform real time vision tasks. Furthermore, to reduce latency and improve real time performance, compression algorithms are proposed and evaluated for streaming real-time video frames to the cloud.

We conduct a feasibility study into the applicability of answer-agnostic question generation models to textbook passages. We show that a significant portion of errors in such systems arise from asking irrelevant or uninterpretable questions and that such errors can be ameliorated by providing summarized input. We find that giving these models human-written summaries instead of the original text results in a significant increase in acceptability of generated questions (33% → 83%) as determined by expert annotators. We also find that, in the absence of human-written summaries, automatic summarization can serve as a good middle ground.

No abstract