Within the field of cognitive neuroscience, Natalia completed her PhD in 2011 with a research that aimed to tackle complex relationships of brain oscillations and their function in efficient information transfer. She then worked as a scientist at University of Leicester (UK) and SISSA (Italy), consolidating her experience in building up multidisciplinary scientific projects. During this period she studied the neural basis of abstract representation, she explored the phenomenon of neural networks optimization to visual input statistics, and co-authored the creation of a forefront technology for online 3D animal movement tracker.
After her research life she started the data scientist career path working at a Tech company in London. She was commissioned to develop, for a supply and treatment utility company, a tool that processes Twitter texts to detect incident-related information and the location associated with it. In October 2019 she joined the BI Team of ProntoPro-Milan, a fast growing marketplace company. There she develops machine learning business solutions for the different departments. She also works with the Product development squad taking care of AB tests analysis and all data analysis and interpretation that is product oriented.
Technical Vision Talk: “Machine Learning based early warning system with social media data”
Ten years ago all complaints to a company came through phone calls. Nowadays all complaints come through social media. This is a revolutionary change, and companies invest a lot of resources on attending these channels. But social media inquiries are massive traced conversations that everybody can listen to. Still, most companies keep treating them as one-to-one phone calls, through monotonous yet inefficient tasks, representing a risk to reputation and customer satisfaction.
Yorkshire Water, the water supply and treatment utility company, servicing more than 5 millions people in a vast area of the north England has identified the need of a tool to bridge the gap between social teams and the reactivity of the incident response team: increasing amount of people communicate for water service issues through social media and the identification of incidents among the massive amount of messages before they escalate represented a challenge for the operators.
Using a series of data science tools as clustering algorithm, a trained neural network for Name Entity Recognition (NER), and a map matching tool to create a geographic and temporal map of all ‘incident specific’ posts, I built an early warning system that allows the company to live-localize incidents, detect the entity of it, and manage efficiently social media complaints using as source of information Twitters.
__________________
Fri. Oct 1 | 9:50 am – Technical Vision Talk: “Machine Learning based early warning system with social media data””.