Advanced search
Start date

Urban insights: deep learning applied to governance in cities

Grant number: 17/16583-6
Support type:Research Grants - Innovative Research in Small Business - PIPE
Duration: October 01, 2018 - September 30, 2019
Field of knowledge:Physical Sciences and Mathematics - Computer Science - Computer Systems
Principal Investigator:Daniel Abujabra Merege
Grantee:Daniel Abujabra Merege
Company:DRSC Soluções Tecnológicas Inteligentes Ltda
CNAE: Desenvolvimento de programas de computador sob encomenda
Portais, provedores de conteúdo e outros serviços de informação na internet
City: São Paulo
Co-Principal Investigators:Rafael Pillon Almeida ; Ricardo Igor Souto Guimarães
Assoc. researchers: Rodrigo Marotti Togneri
Associated scholarship(s):18/22126-0 - Urban Insights: deep learning applied to governance in cities, BP.TT
18/22123-0 - Urban Insights: deep learning applied to governance in cities, BP.TT
18/21776-0 - Urban insights: deep learning applied to governance in cities, BP.PIPE


Cities worldwide need to be developed technologically and to foster innovation to provide better services to citizens. According to the United Nations, people living in urban areas over the world are expected to account for more than 60% of the world's population by 2050 (UNITED NATIONS, 2014), which places greater pressure on public services provided by cities. The data produced by the population and the operation of public services is essential to have a holistic and complete view of the city, raising public governance to a higher level. One of the technology field that has gained attention recently for analysis of large amount of data generated by billions of devices connected to the global computer network is Artificial Intelligence. Machine Learning systems can be used to identify objects in images, transform speeches into texts, improve search systems for relevant result, and so forth; Natural Language Processing (NLP) activities, especially classification, sentiment analysis and language translation are highlighted out of this group. The inflection point in the curve of adoption of algorithms of Machine Learning in real applications happened with the emergence of Deep Learning techniques. In the past, the development of this type of system required specialized technical knowledge in this area, so that learning systems, generally classifiers, could identify patterns from pure data passed as inputs (LECUN et al., 2015). With the advent of Deep Learning techniques - through which multiple layers of representation transform pure data into larger and more abstract representations -, Artificial Intelligence has been applied in different areas of Science, business and government. However, although there have been advances in the application of Artificial Intelligence methods to societal problems, there is a strong lack of technologies aimed at the public sector, especially when it comes to public governance. This project aims to develop an intelligence platform for urban management using Deep Learning techniques. Throughout the development of the project, architectures of a deep neural network will be developed and tested to create an interpreter that will extract relevant information about urban problems from texts coming from a variety of sources (e.g. social media, newspapers, magazines, blogs, urban sensors), and will classify them into urban categories, so that this information is provided on charts and tables in an analytical panel called Urban Insights. As a result of the exploratory phase of the project, it is expected to achieve the state-of-the-art accuracy of deep neural network in text classification task, comparing the results with related works (CONNEAU, 2016), (PORIA, CAMBRIA and GELBUKH, 2016) and (WEHRMANN et al., 2017). Afterwards, with the information and visualization being validated by public managers, we expect to transform the Panel into a marketable product for city governments and for companies interested in urban data (such as insurers, for example), to bring efficiency to the in-depth analysis of urban problems and to reach rapid and innovative responses to these problems, coming not only from municipalities but also from the entire urban innovation ecosystem. (AU)