Search

Natural language generation (NLG) is a critical component in conversational systems, owing to its role of formulating a correct and natural text response. Traditionally, NLG components have been deployed using template-based solutions. Although neural network solutions recently developed in the research community have been shown to provide several benefits, deployment of such model-based solutions has been challenging due to high latency, correctness issues, and high data needs. In this paper, we present approaches that have helped us deploy data-efficient neural solutions for NLG in conversational systems to production. We describe a family of sampling and modeling techniques to attain production quality with light-weight neural network models using only a fraction of the data that would be necessary otherwise, and show a thorough comparison between each. Our results show that domain complexity dictates the appropriate approach to achieve high data efficiency. Finally, we distill the lessons from our experimental findings into a list of best practices for production-level NLG model development, and present them in a brief runbook. Importantly, the end products of all of the techniques are small sequence-to-sequence models (2Mb) that we can reliably deploy in production., Accepted for publication in COLING 2020

We studied the paracrine function of mesenchymal stem cells (MSCs) derived from various sources in response to pulsed focused ultrasound (pFUS). Human adipose tissue (AD), bone marrow (BM), and umbilical cord (UC) derived MSCs were exposed to pFUS at two intensities: 0.45 W/cm2 ISATA (310 kPa PNP) and 1.3 W/cm2 ISATA (540 kPa PNP). Following pFUS, the viability and proliferation of MSCs were assessed using a hemocytometer and confocal microscopy, and their secreted cytokine profile determined using a multiplex ELISA. Our findings showed that pFUS can stimulate the production of immunomodulatory, anti-inflammatory, and angiogenic cytokines from MSCs which was dependent on both the source of MSC being studied and the acoustic intensity employed. These important findings set the foundation for additional mechanistic and validation studies using this novel noninvasive and clinically translatable technology for modulating MSC biology.

Genetically encoded voltage indicators (GEVIs) are emerging optical tools for acquiring brain-wide cell-type-specific functional data at unparalleled temporal resolution. To broaden the application of GEVIs in high-speed multispectral imaging, we used a high-throughput strategy to develop voltage-activated red neuronal activity monitor (VARNAM), a fusion of the fast Acetabularia opsin and the bright red fluorophore mRuby3. Imageable under the modest illumination intensities required by bright green probes (

PubMed: 30740513, Neutrophils have a critical role in regulating the immune system. The immune system is compromised during chemotherapy, increasing infection risks and imposing a need for regular monitoring of neutrophil counts. Although commercial hematology analyzers are currently used in clinical practice for neutrophil counts, they are only available in clinics and hospitals, use large blood volumes, and are not available at the point of care (POC). Additionally, phlebotomy and blood processing require trained personnel, where patients are often admitted to hospitals when the infections are at late stage due to lack of frequent monitoring. Here, a reliable method is presented that selectively captures and quantifies white blood cells (WBCs) and neutrophils from a finger prick volume of whole blood by integrating microfluidics with high-resolution imaging algorithms. The platform is compact, portable, and easy to use. It captures and quantifies WBCs and neutrophils with high efficiency (> 95%) and specificity (> 95%) with an overall 4.2% bias compared to standard testing. The results from a small cohort of patients (N = 11 healthy, N = 5 lung and kidney cancer) present a unique disposable cell counter, demonstrating the ability of this tool to monitor neutrophil and WBC counts within clinical or in resource-constrained environments.