SHEDR: An End-to-End Deep Neural Event Detection and Recommendation Framework for Hyperlocal News Using Social Media

Residents often rely on newspapers and television to gather hyperlocal news for community awareness and engagement. More recently, social media have emerged as an increasingly important source of hyperlocal news. Thus far, the literature on using social media to create desirable societal benefits, such as civic awareness and engagement, is still in its infancy. One key challenge in this research stream is to timely and accurately distill information from noisy social media data streams to community members. In this work, we develop SHEDR (social media–based hyperlocal event detection and recommendation), an end-to-end neural event detection and recommendation framework with a particular use case for Twitter to facilitate residents’ information seeking of hyperlocal events. The key model innovation in SHEDR lies in the design of the hyperlocal event detector and the event recommender. First, we harness the power of two popular deep neural network models, the convolutional neural network (CNN) and long short-term memory (LSTM), in a novel joint CNN-LSTM model to characterize spatiotemporal dependencies for capturing unusualness in a region of interest, which is classified as a hyperlocal event. Next, we develop a neural pairwise ranking algorithm for recommending detected hyperlocal events to residents based on their interests. To alleviate the sparsity issue and improve personalization, our algorithm incorporates several types of contextual information covering topic, social, and geographical proximities. We perform comprehensive evaluations based on two large-scale data sets comprising geotagged tweets covering Seattle and Chicago. We demonstrate the effectiveness of our framework in comparison with several state-of-the-art approaches. We show that our hyperlocal event detection and recommendation models consistently and significantly outperform other approaches in terms of precision, recall, and F-1 scores. Summary of Contribution: In this paper, we focus on a novel and important, yet largely underexplored application of computing—how to improve civic engagement in local neighborhoods via local news sharing and consumption based on social media feeds. To address this question, we propose two new computational and data-driven methods: (1) a deep learning–based hyperlocal event detection algorithm that scans spatially and temporally to detect hyperlocal events from geotagged Twitter feeds; and (2) A personalized deep learning–based hyperlocal event recommender system that systematically integrates several contextual cues such as topical, geographical, and social proximity to recommend the detected hyperlocal events to potential users. We conduct a series of experiments to examine our proposed models. The outcomes demonstrate that our algorithms are significantly better than the state-of-the-art models and can provide users with more relevant information about the local neighborhoods that they live in, which in turn may boost their community engagement.

Early Versus Late Mild Cognitive Impairment: Neural Event-Related Oscillations During a Go/No Go Task

Amnestic mild cognitive impairment (aMCI) is marked by episodic memory deficits, which is used to classify individuals into early MCI (EMCI) and late MCI (LMCI). Growing evidence suggests that individuals with EMCI and LMCI differ in other cognitive functions including cognitive control, but these are less frequently studied. Using a semantic Go/NoGo task, we examined differences in cognitive control between EMCI and LMCI on behavioral (accuracy and reaction time) and neural (scalp-recorded event-related oscillations in theta and alpha band) measures. Although no behavioral differences were observed between the groups, EMCI and LMCI groups differed in patterns of neural oscillations for Go compared to NoGo trials. The EMCI group showed differences in theta power at central electrodes and alpha power at central and centro-parietal electrodes between Go and NoGo trials, while the LMCI group did not exhibit such differences. Furthermore, the LMCI group had higher theta synchronization on Go trials at central electrodes compared to the EMCI group. These findings suggest that while behavioral differences may not be observable, neural changes underlying cognitive control processes may differentiate EMCI and LMCI stages and may be useful to understand the trajectory of aMCI.

Aura Mapping: Where Vision and Somatosensation Meet

While migraine auras are most frequently visual, somatosensory auras are also relatively common. Both are characterized by the spread of activation across a cortical region containing a spatial mapping of the sensory (retinal or skin) surface. When both aura types occur within a single migraine episode, they may offer an insight into the neural mechanism which underlies them. Could they both be initiated by a single neural event, or do the timing and laterality relationships between them demand multiple triggers? The observations reported here were carried out 25 years ago by a group of six individuals with migraine with aura. They timed, described and mapped their visual and somatosensory auras as they were in progress. Twenty-nine episode reports are summarized here. The temporal relationship between the onset of the two auras was quite variable within and across participants. Various forms of the cortical spreading depression hypothesis of migraine aura are evaluated in terms of whether they can account for the timing, pattern of symptom spread and laterality of the recorded auras.

Neural event segmentation of continuous experience in human infants

Although sensory input is continuous, we perceive and remember discrete events. Event segmentation has been studied extensively in adults, but little is known about how the youngest minds experience the world. The main impediment to studying event segmentation in infants has been a reliance on explicit parsing tasks that are not possible at this age. fMRI has recently proven successful at measuring adult event segmentation during task-free, naturalistic perception. Applied to infants, this could reveal the nature of their event segmentation, from low-level sensory transients to high-level cognitive boundaries. We collected fMRI data from 25 adults and 25 infants less than one year of age watching the same short movie. Neural events were defined by the stability of voxel activity patterns. In adults, we replicated a hierarchical gradient of event timescales, from shorter events in early visual regions to longer events in later visual and narrative regions. In infants, however, longer events were found throughout the brain, including in a second dataset. Infant event structure fit adult data and vice versa, but adult behavioral boundaries were differently expressed in adult and infant brains. These findings have implications for the nature of infant experience and cognition.

Neural Event Semantics for Grounded Language Understanding

We present a new conjunctivist framework, neural event semantics (NES), for compositional grounded language understanding. Our approach treats all words as classifiers that compose to form a sentence meaning by multiplying output scores. These classifiers apply to spatial regions (events) and NES derives its semantic structure from language by routing events to different classifier argument inputs via soft attention. NES is trainable end-to-end by gradient descent with minimal supervision. We evaluate our method on compositional grounded language tasks in controlled synthetic and real-world settings. NES offers stronger generalization capability than standard function-based compositional frameworks, while improving accuracy over state-of-the-art neural methods on real-world language tasks.