As human beings utilize computing technologies to mediate multiple aspects of their lives, cyberbullying has grown as an important societal challenge. Cyberbullying may lead deep psychiatric and emotional disorders for those affected. Hence, there is urgent need devise automated methods detection prevention. While recent efforts have defined sophisticated text processing detection, are yet few that leverage visual data automatically detect cyberbullying. Based on early analysis a public,...

Small footprint embedded devices require keyword spotters (KWS) with small model size and detection latency for enabling voice assistants. Such a is often referred to as \textit{wake word} it used wake up assistant enabled devices. Together word detection, accurate estimation of endpoints (start end) an important task KWS. In this paper, we propose two new methods detecting the words in neural KWS that use single-stage word-level networks. Our results show techniques give superior accuracy...

In this work, we propose Tiny-CRNN (Tiny Convolutional Recurrent Neural Network) models applied to the problem of wakeword detection, and augment them with scaled dot product attention. We find that, compared Network models, False Accepts in a 250k parameter budget can be reduced by 25% 10% reduction size using based on architecture, get up 32% at 50k 75% word-level Dense models. discuss solutions challenging performing inference streaming audio as well differences start-end index errors...

In this work, we propose small footprint Convolutional Recurrent Neural Network models applied to the problem of wakeword detection and augment them with scaled dot product attention. We find that false accepts compared in a 250k parameter budget can be reduced by 25% 10% reduction size using CRNNs, get up 32% improvement at 50k 75% word-level Dense models. discuss solutions challenging performing inference on streaming audio as well differences start-end index errors latency comparison CNN,...

The COVID-19 pandemic has quickly disrupted global trade and travel, affecting our day-to-day lives. In its 209th report, released on August 16, 2020, World Health Organization (WHO) reported that the acute respiratory syndrome (SARS-CoV2) coronavirus illness (COVID-19) killed more than 379,941 people infected 6000000 worldwide. custom of donning a safety mask evolved. Many public service providers may soon need their clients to wear necessary masks in order use services. serve worldwide...

We propose a novel approach for semi-supervised learning (SSL) designed to overcome distribution shifts between training and real-world data arising in the keyword spotting (KWS) task. Shifts from are key challenge KWS tasks: when new model is deployed on device, gating of accepted undergoes shift distribution, making problem timely updates via subsequent deployments hard. Despite shift, we assume that marginal distributions labels do not change. utilize modified teacher/student framework,...

Conversational agents commonly utilize keyword spotting (KWS) to initiate voice interaction with the user.For user experience and privacy considerations, existing approaches KWS largely focus on accuracy, which can often come at expense of introduced latency.To address this tradeoff, we propose a novel approach control model latency generalizes any loss function without explicit knowledge endpoint.Through single, tunable hyperparameter, our enables one balance detection accuracy for targeted...

In this work, we propose Tiny-CRNN (Tiny Convolutional Recurrent Neural Network) models applied to the problem of wakeword detection, and augment them with scaled dot product attention. We find that, compared Network models, False Accepts in a 250k parameter budget can be reduced by 25% 10% reduction size using based on architecture, get up 32% at 50k 75% word-level Dense models. discuss solutions challenging performing inference streaming audio as well differences start-end index errors...