No abstract

Social media data represent an important resource for behavioral analysis of the ageing population. This paper addresses the problem of age prediction from Twitter dataset, where the prediction issue is viewed as a classification task. For this purpose, an innovative model based on Convolutional Neural Network is devised. To this end, we rely on languagerelated features and social media specific metadata. More specifically, we introduce two features that have not been previously considered in the literature: the content of URLs and hashtags appearing in tweets. We also employ distributed representations of words and phrases present in tweets, hashtags and URLs, pre-trained on appropriate corpora in order to exploit their semantic information in age prediction. We show that our CNN-based classifier, when compared with an SVM baseline model, yields an improvement of 12.3% and 6.6% in the micro-averaged F1 score on the Dutch and English datasets, respectively.

Prevalent weather prediction methods are based on sensor data, collected by satellites and a sparse grid of stationary weather stations. Various initiatives improve the prediction models by including additional data sources such as mobile weather sensors, mobile phones, and micro weather stations of, for example, smart homes. The underlying computing paradigm is predominantly centralized, with all data collected and analyzed in the cloud. This solution is not scalable. When the spatial and temporal density of weather sensor data grows, the required data transmission capacities and computational resources become unfeasible. We identify the challenges posed by spatial distribution of a weather prediction model, and suggest solutions for those challenges. We propose EDISON: an edgenative interpolation approach based on AI methods, distributed horizontally on edge servers. Finally, we demonstrate EDISON with a simple, simulated setup.

Due to its asynchronous message-sharing and real-time capabilities, Twitter offers a valuable opportunity to detect events in a timely manner. Existing approaches for event detection have mainly focused on building a temporal profile of named entities and detecting unusually large bursts in their usage to signify an event. We extend this line of research by incorporating external knowledge bases such as DBPedia, WordNet; and exploiting specific features of Twitter for efficient event detection. We show that our system utilizing temporal, social, and Twitter-specific features yields improvement in the precision, recall, and DERate on the benchmarked Events2012 corpus compared to the stateof-the-art approaches.

Inferring demographic intelligence from unlabeled social media data is an actively growing area of research, challenged by low availability of ground truth annotated training corpora. High-accuracy approaches for labeling demographic traits of social media users employ various heuristics that do not scale up and often discount non-English texts and marginalized users. First, we present a framework for inferring the demographic attributes of Twitter users from their profile pictures (avatars) using the Microsoft Azure Face API. Second, we measure the inter-rater agreement between annotations made using our framework against two pre-labeled samples of Twitter users (N1=1163; N2=659) whose age labels were manually annotated. Our results indicate that the strength of the inter-rater agreement (Gwet's AC1=0.89; 0.90) between the gold standard and our approach is 'very good' for labelling the age group of users. The paper provides a use case of Computer Vision for enabling the development of large cross-sectional labeled datasets, and further advances novel solutions in the field of demographic inference from short social media texts.