Search for: All records

The outbreak of the novel coronavirus, COVID-19, has become one of the most severe pandemics in human history. In this paper, we propose to leverage social media users as social sensors to simultaneously predict the pandemic trends and suggest potential risk factors for public health experts to understand spread situations and recommend proper interventions. More precisely, we develop novel deep learning models to recognize important entities and their relations over time, thereby establishing dynamic heterogeneous graphs to describe the observations of social media users. A dynamic graph neural network model can then forecast the trends (e.g. newly diagnosed cases and death rates) and identify high-risk events from social media. Based on the proposed computational method, we also develop a web-based system for domain experts without any computer science background to easily interact with. We conduct extensive experiments on large-scale datasets of COVID-19 related tweets provided by Twitter, which show that our method can precisely predict the new cases and death rates. We also demonstrate the robustness of our web-based pandemic surveillance system and its ability to retrieve essential knowledge and derive accurate predictions across a variety of circumstances. Our system is also available at http://scaiweb.cs.ucla.edu/covidsurveiller/ . This article is part of the theme issue ‘Data science approachs to infectious disease surveillance’. 

Adversarial attacks against machine learning models have threatened various real-world applications such as spam filtering and sentiment analysis. In this paper, we propose a novel framework, learning to discriminate perturbations (DISP), to identify and adjust malicious perturbations, thereby blocking adversarial attacks for text classification models. To identify adversarial attacks, a perturbation discriminator validates how likely a token in the text is perturbed and provides a set of potential perturbations. For each potential perturbation, an embedding estimator learns to restore the embedding of the original word based on the context and a replacement token is chosen based on approximate kNN search. DISP can block adversarial attacks for any NLP model without modifying the model structure or training procedure. Extensive experiments on two benchmark datasets demonstrate that DISP significantly outperforms baseline methods in blocking adversarial attacks for text classification. In addition, in-depth analysis shows the robustness of DISP across different situations. 

Stress is a common problem in modern life that can bring both psychological and physical disorder. Wearable sensors are commonly used to study the relationship between physical records and mental status. Although sensor data generated by wearable devices provides an opportunity to identify stress in people for predictive medicine, in practice, the data are typically complicated and vague and also often fragmented. In this paper, we propose DataCompletion with Diurnal Regularizers (DCDR) and TemporallyHierarchical Attention Network (THAN) to address the fragmented data issue and predict human stress level with recovered sensor data. We model fragmentation as a sparsity issue. The nuclear norm minimization method based on the low-rank assumption is first applied to derive unobserved sensor data with diurnal patterns of human behaviors. A hierarchical recurrent neural network with the attention mechanism then models temporally structural information in the reconstructed sensor data, thereby inferring the predicted stress level. Data for this study were from 75 undergraduate students (taken from a sample of a larger study) who provided sensor data from smart wristbands. They also completed weekly stress surveys as ground-truth labels about their stress levels. This survey lasted 12 weeks and the sensor records are also in this period. The experimental results demonstrate that our approach significantly outperforms conventional methods in both data completion and stress level prediction. Moreover, an in-depth analysis further shows the effectiveness and robustness of our approach.