Infant Mortality Prediction
This work was done in collaboration with the Heller School for Social Policy and Management (Brandeis University). We apply several machine learning models in data extracted from birth certificates to show how well they can perform at the infant mortality prediction problem. We focus on socio-economic factors, and show that they can improve the overall performance of the prediction task, while not requiring data from future visits to the doctor. We presented part of this project during the KDD Health Day 2019 [paper][code]. The work has now been accepted in the Data Mining and Knowledge Discovery journal. Journal paper, code and data will be published here soon! (update: Dec. 2020)

News Monitor: A Framework for News Monitoring and Analysis
This work is part of our efforts at the KDDLab to build a News Monitoring system that collects news articles from several RSS feeds and combines them with related posts from social networks. The system then analyzes the data and performs a set of tasks, such as: first story detection, opinion mining, knowledge base construction, and question answering. Paper, videos, code and demo will be published here soon! (update: Dec. 2020)

Ranking Notable News using Knowledge Graph Stuctures
This projest is part of the work I did while I was a research intern at the Graph Analytics team of Bloomberg AI in London, advised by Edgar Meij and Giorgio Stefanoni. We show how the structured representations of news stories can be used to rank them based on their notability, and we use a crowdsourcing task to obtain golden truth labels. This work is published in ECIR 2020 [paper][video], and was also presented in the MEDIATE workshop, ICWSM 2020 [video].

Sub-Event Detection in Content Graphs
In this project, we propose to detect events in a Content Graph, and we summarize each of them in a timeline of highlights. We are one of the first ones to define the sub-event detection problem, and we conduct an evaluation on a manually annotated dataset from Twitter that contains football games from Premier League. This work is published in ICDE 2018 [paper] [code].

Event Detection in Social Networks
In this project, we define a Content Network, we show how to reveal hidden links in this graph and we use this structure to detect events. We show how our method can also be used in an event ranking setup. We evaluate our method using other baselines and the state-of-the-art in a Twitter dataset that we have manually annotated. In addition to the above, this method has applications in trend detection on co-authorship and product-review graphs. This work is published in CIKM 2017 [paper] [code].