Search

Complex natural language understanding modules in dialog systems have a richer understanding of user utterances, and thus are critical in providing a better user experience. However, these models are often created from scratch, for specific clients and use cases, and require the annotation of large datasets. This encourages the sharing of annotated data across multiple clients. To facilitate this we introduce the idea of intent features: domain and topic agnostic properties of intents that can be learned from the syntactic cues only, and hence can be shared. We introduce a new neural network architecture, the Global-Local model, that shows significant improvement over strong baselines for identifying these features in a deployed, multi-intent natural language understanding module, and, more generally, in a classification setting where a part of an utterance has to be classified utilizing the whole context., Camera-ready for NAACL 2021 Industry Track

Many tasks in natural language processing, such as named entity recognition and slot-filling, involve identifying and labeling specific spans of text. In order to leverage common models, these tasks are often recast as sequence labeling tasks. Each token is given a label and these labels are prefixed with special tokens such as B- or I-. After a model assigns labels to each token, these prefixes are used to group the tokens into spans. Properly parsing these annotations is critical for producing fair and comparable metrics; however, despite its importance, there is not an easy-to-use, standardized, programmatically integratable library to help work with span labeling. To remedy this, we introduce our open-source library, iobes. iobes is used for parsing, converting, and processing spans represented as token level decisions.

Because large, human-annotated datasets suffer from labeling errors, it is crucial to be able to train deep neural networks in the presence of label noise. While training image classification models with label noise have received much attention, training text classification models have not. In this paper, we propose an approach to training deep networks that is robust to label noise. This approach introduces a non-linear processing layer (noise model) that models the statistics of the label noise into a convolutional neural network (CNN) architecture. The noise model and the CNN weights are learned jointly from noisy training data, which prevents the model from overfitting to erroneous labels. Through extensive experiments on several text classification datasets, we show that this approach enables the CNN to learn better sentence representations and is robust even to extreme label noise. We find that proper initialization and regularization of this noise model is critical. Further, by contrast to results focusing on large batch sizes for mitigating label noise for image classification, we find that altering the batch size does not have much effect on classification performance.

Interference experiments provide a simple yet powerful tool to unravel fundamental features of quantum physics. Here we engineer an RF-driven, time-dependent bilinear coupling that can be tuned to implement a robust 50:50 beamsplitter between stationary states stored in two superconducting cavities in a three-dimensional architecture. With this, we realize high contrast Hong-Ou- Mandel (HOM) interference between two spectrally-detuned stationary modes. We demonstrate that this coupling provides an efficient method for measuring the quantum state overlap between arbitrary states of the two cavities. Finally, we showcase concatenated beamsplitters and differential phase shifters to implement cascaded Mach-Zehnder interferometers, which can control the signature of the two-photon interference on-demand. Our results pave the way toward implementation of scalable boson sampling, the application of linear optical quantum computing (LOQC) protocols in the microwave domain, and quantum algorithms between long-lived bosonic memories., Comment: 6 pages, 5 figures

We introduce Baseline: a library for reproducible deep learning research and fast model development for NLP. The library provides easily extensible abstractions and implementations for data loading, model development, training and export of deep learning architectures. It also provides implementations for simple, high-performance, deep learning models for various NLP tasks, against which newly developed models can be compared. Deep learning experiments are hard to reproduce, Baseline provides functionalities to track them. The goal is to allow a researcher to focus on model development, delegating the repetitive tasks to the library.

Background: Nevus comedonicus (NC) is rarely associated with the histopathologic pattern of follicular epidermolytic hyperkeratosis (EHK). We found eight cases reported. In one case, the condition was transmitted to the offspring in the form of generalized EHK. Methods: We describe a case of linear NC with EHK in a 46-year-old woman. Results: Histopathologic examination revealed the typical features of NC. Additionally, the follicular epithelial walls showed EHK with characteristic perinuclear vacuolization in the stratum spinosum and stratum granulosum and large, irregular keratohyalin granules in the granular cell layer. Conclusions: The clinical and histopathologic features of this case are consistent with a diagnosis of linear NC with EHK. Lesions of NC suspected by clinical exam should be examined microscopically to look for features of EHK. If present, patients should be educated about the risk, albeit rare, of passing on a more severe form of the disorder to subsequent generations.

Following their discovery in the early 60's, there has been a continuous quest for ways to take advantage of the extraordinary properties of shape memory alloys (SMAs). These intermetallic alloys can be extremely compliant while retaining the strength of metals and can convert thermal energy to mechanical work. The unique properties of SMAs result from a reversible difussionless solid-to-solid phase transformation from austenite to martensite. The integration of SMAs into composite structures has resulted in many benefits, which include actuation, vibration control, damping, sensing, and self-healing. However, despite substantial research in this area, a comparable adoption of SMA composites by industry has not yet been realized. This discrepancy between academic research and commercial interest is largely associated with the material complexity that includes strong thermomechanical coupling, large inelastic deformations, and variable thermoelastic properties. Nonetheless, as SMAs are becoming increasingly accepted in engineering applications, a similar trend for SMA composites is expected in aerospace, automotive, and energy conversion and storage related applications. In an effort to aid in this endeavor, a comprehensive overview of advances with regard to SMA composites and devices utilizing them is pursued in this paper. Emphasis is placed on identifying the characteristic responses and properties of these material systems as well as on comparing the various modeling methodologies for describing their response. Furthermore, the paper concludes with a discussion of future research efforts that may have the greatest impact on promoting the development of SMA composites and their implementation in multifunctional structures.; International audience; Following their discovery in the early 60's, there has been a continuous quest for ways to take advantage of the extraordinary properties of shape memory alloys (SMAs). These intermetallic alloys can be extremely compliant while retaining the strength of metals and can convert thermal energy to mechanical work. The unique properties of SMAs result from a reversible difussionless solid-to-solid phase transformation from austenite to martensite. The integration of SMAs into composite structures has resulted in many benefits, which include actuation, vibration control, damping, sensing, and self-healing. However, despite substantial research in this area, a comparable adoption of SMA composites by industry has not yet been realized. This discrepancy between academic research and commercial interest is largely associated with the material complexity that includes strong thermomechanical coupling, large inelastic deformations, and variable thermoelastic properties. Nonetheless, as SMAs are becoming increasingly accepted in engineering applications, a similar trend for SMA composites is expected in aerospace, automotive, and energy conversion and storage related applications. In an effort to aid in this endeavor, a comprehensive overview of advances with regard to SMA composites and devices utilizing them is pursued in this paper. Emphasis is placed on identifying the characteristic responses and properties of these material systems as well as on comparing the various modeling methodologies for describing their response. Furthermore, the paper concludes with a discussion of future research efforts that may have the greatest impact on promoting the development of SMA composites and their implementation in multifunctional structures.

We demonstrate imaging of neutral atoms via the light scattered during continuous Raman sideband cooling. We detect single atoms trapped in optical tweezers while maintaining a significant motional ground-state fraction. The techniques presented provide a framework for single-atom resolved imaging of a broad class of atomic species.

Nephrogenic fibrosing dermopathy (NFD) and calciphylaxis are rare conditions that are associated with chronic kidney disease. Histopathologic changes, including dystrophic dermal calcification, often in association with elastic fibers have been observed in NFD and calciphylaxis. A pattern of dermal elastic fiber calcification that mimics pseudoxanthoma elasticum (PXE) has been previously reported as an incidental finding in the setting of calciphylaxis. Despite a shared association with renal disease and abnormal calcium deposits, however, NFD and calciphylaxis are discrete pathologic processes with distinct clinical and histopathologic features. Criteria for each are reviewed through case presentation of a patient meeting the clinical and histopathologic criteria for both NFD and calciphylaxis with histologic features mimicking PXE.

Background Our objective was to examine factors associated with distributive injection equipment sharing and how needle exchange programs (NEPs) can help reduce distributive sharing among injection drug users (IDUs). Methods 145 English speaking Canadian IDUs ages 16 years and over who had injected in the past 30 days were recruited for a cross-sectional survey. Participants were asked about their socio-demographic characteristics, HIV risk behaviours, social support, drug treatment readiness, program satisfaction, health and social service use and NEP drug use. Bivariate statistics and logistic regression were used to characterize the population and examine correlates of sharing behaviour. Results More IDUs reported distributive sharing of cookers (45%) than needles (36%) or other types of equipment (water 36%; filters 29%; swabs 8%). Regression analyses revealed the following factors associated with distributing used cookers: a history of cocaine/crack injection, an Addiction Severity Index (ASI) score indicative of a mental health problem, and older than 30 years of age. Factors associated with giving away used water included: male, injected methadone, injected other stimulants and moved 3+ times in the past 6 months. Factors associated with giving away used filters included: injected cocaine/crack or stayed overnight on the street or other public place. Factors associated with giving away swabs included: an ASI mental health score indicative of a mental health problem, and HCV negative status. Conclusions Our findings show that more IDUs give away cookers than needles or other injection equipment. While the results showed that correlates of sharing differed by piece of equipment, each point to distributive sharing by the most marginalized IDUs. Targeting prevention efforts to reduce equipment sharing in general, and cookers in particular is warranted to reduce use of contaminated equipment and viral transmission.