Link a PDF con los proyectos en español
Comunicado de prensa en español

The Google Research Awards for Latin America aim to identify and support students and professors who are pursuing cutting-edge research in specific fields related to Computer Science. They began in Brazil in 2013 as an initiative developed by Google’s Center of Engineering in Belo Horizonte. Last year, they were expanded to the rest of Latin America, with a budget of USD $300,000 distributed among 24 researchers. For this third edition, Google is providing USD $600,000, thus doubling the number of supported grantees to 48 researchers.


The areas supported in this year’s awards are:


  • Geo/maps

  • Human-computer interaction

  • Information retrieval, extraction & organization (including semantic graphs)

  • Internet of things (including smart cities)

  • Machine learning and data mining

  • Mobile

  • Natural language processing

  • Physical interfaces and immersive experiences

  • Privacy

  • Other topics related to web research


Winners

The winning projects for the 2016 edition come from the following countries: Argentina (1), Brazil (17), Chile (2), Colombia (1), Mexico (2) and Peru (1). Out of the 24 winning projects, there are 7 extensions to previous - 2015 - winners, in Brazil (4), Chile (1), Colombia (1) and Mexico (1).


The Awards include monthly grants that are paid to both the student and the advisor professor for the project. PhD students receive USD$ 1,200 per month, while their advisors receive USD 750 per month. Master's students receive USD$ 750 monthly, while their professors receive USD$ 675 monthly. The grants are awarded for a period of one year with the possibility of annual renewal up to a limit of two years in total for master students and three years for PhD students.


Argentina

Title: Prognosis in a Box: Computational Characterization of Mental States

Universidad de Buenos Aires, AR

Professor and student: Diego Fernandez Slezak; Facundo Carrillo

Summary: The focus of this project is to develop a tool to assist psychiatric practice. The proposed model would capture markers of psychiatric and neurological conditions to be integrated in a cognitive computing system.


Brazil

Title: Low-cost Autonomous Stations for Measuring the Impact of Climate Change on Glaciers

Universidade Federal do Rio Grande, BR

Professor and student: Jorge Arigony-Neto; Guilherme Tomaschewski Netto

Summary: This project proposes new methods for monitoring glaciers by means of a network of low-cost open source equipments for measuring weather parameters and glacier melting directly on glacier surfaces with remote transfer of data.


Title: Automatic Generation of Video Synopsis with Convolutional and Recurrent Neural Networks

Pontifícia Universidade Católica do Rio Grande do Sul, BR

Professor and student: Rodrigo Coelho Barros; Jônatas Wehrmann

Summary: This proposal defines the initial guidelines for solving the problem of automatically generating synopses from videos.


Title: Automatic Reasoning about Contractual Clauses

Pontifícia Universidade Católica do Rio Grande do Sul, BR

Professor and student: Felipe Meneguzzi; João Paulo Aires

Summary: This project aims to develop an approach for reasoning about complex norms in natural language documents (typically contracts) and detecting norm conflicts and contract inconsistencies.


Title: Multimodal Assistive Domotics Including Augmentative and Alternative Communication

Universidade Federal do Espírito Santo, BR

Professor and student: Teodiano Freire Bastos-Filho; Alexandre Bissoli

Summary: The goal of this project is to develop a new assistive system that can be used by people with severe motor disabilities to enable them to control home devices and communicate by means of biological signals captured from muscles or eyes.


Title: Using Search Engine Idea to Repurpose Approved Drugs as Antibiotics

Universidade Federal de Minas Gerais, BR

Professor and student: Marcos Augusto dos Santos; Rita Silvério-Machado

Summary: The project will develop a model for the identification of potential antibiotic drugs using a repurposing approach.


Title: Discussion-Based Entity Representation

Universidade Federal de Minas Gerais, BR

Professor and student: Pedro Olmo Stancioli Vaz de Melo; Túlio Corrêa Loures

Summary: The goal is to develop a method for learning entity representations from online discussions that can automatically generate a summary of that entity even when nothing is explicitly known about it, generating attributes for that entity automatically.


Title: Hot Spot Mining from Case-Control Trajectories

Universidade Federal de Minas Gerais, BR

Professor and student: Wagner Meira Junior; Roberto C.S.N.P. Souza

Summary: This project focuses on detecting localized spatial regions or zones, where the probability of some targeted event occurrence is higher than in the rest of the map.


Title: Mapping Graffiti from Google Street View Data

Universidade Federal de Pelotas

Professor and student: Ricardo Matsumura de Araujo; Glauco Roberto Munsberg

Summary: The project aims to use Google Street View data to automatically map graffitis, allowing an unsupervised learning algorithm to cluster graffiti styles and provide useful sociological information on the resulting map.


Title: Automated Generation of Tutorials from Q&A Sites

Universidade Federal de Uberlândia

Professor and student: Marcelo de Almeida Maia; Adriano Mendonça Rocha

Summary: The team’s proposal is aimed at leveraging Q&A sites to automatically produce structured additional documentation for software developers from the crowd knowledge deployed at Q&A sites.


Title: Automated Data­-Driven Screening of Diabetic Retinopathy

Universidade Estadual de Campinas

Professor and student: Anderson Rocha; José Ramon Trindade Pires

Summary: This project aims to gather enough information directly from the available data in order to design a more effective, unified, and less human-­centered classification system for the early detection of Diabetic Retinopathy.


Title: Reliable Automated Melanoma Screening for the Real World

Universidade Estadual de Campinas

Professor and student: Eduardo Alves do Valle Junior; Michel Silva Fornacial

Summary: This research aims to address Melanoma detection by applying cutting-edge approaches for image classification, advancing the cooperation with dermatologists and gathering new images that can be used by the community for further research.


Title: Bringing Down Barriers: Including Blind People in Cooperative Modeling

Universidade de São Paulo

Professor and student: Anarosa Alves Franco Brandão; Leandro Luque

Summary: In this project, the researchers’ aim is to propose and test a method for including blind people in cooperative modeling activities.


Title: Improving Dwell-Free Eye Typing Techniques Using a Noisy Model-Based Approach

Instituto de Ciências Matemáticas e de Computação da Universidade de São Paulo

Professor and student: Maria da Graça Campos Pimentel; Raíza T. S. Hanada

Summary: In this proposal, the team intends to improve the eye gaze-based text entry systems by adopting a noisy channel approach, based on error models, to achieve faster user entry rate and reduce eye fatigue due to typing errors.


Title: Optimizing Ensembles of Boosted Additive Bagged Trees for Learning-to-Rank. EXTENSION from 2015

Universidade Federal de Minas Gerais

Professor and student: Marcos André Gonçalves; Clebson C. A. de Sá

Summary: In 2015, this research focused on solving the problem of the information retrieval known as “Learning to Rank (L2R)”, the primary goal of which was to learn how to combine predictors to leverage the performance of ranking models, by combining in an effective manner two ensemble-based Machine Learning (ML) strategies: Boosting and Bagging. For this year's research, the focus will be on covering a lot of ground regarding ML approaches.


Title: Improving Deep Reinforcement Learning through Knowledge Transfer. EXTENSION from 2015

Universidade de São Paulo

Professor and student: Anna Helena Reali Costa; Ruben Glatt

Summary: The goal of this project is to develop intelligent Reinforcement Learning (RL) agents, which are able to solve different tasks more efficiently with the use of Transfer Learning (TL) approaches. A remaining key question to be addressed this year is how to determine the similarity between tasks automatically, without further domain knowledge.


Title: Beyond Relevance: Addressing Novelty, Diversity and Personalization in Tag Recommendation. EXTENSION from 2015

Universidade Federal de Minas Gerais

Professor and student: Jussara Marques de Almeida; Fabiano Muniz Belém

Summary: In 2015, this project focused on developing novel tag recommendation solutions that effectively address multiple aspects related to the problem, notably, relevance, novelty, diversity and personalization to the interests of the user. For this year's project, the goal is to extend the research by exploring new opportunities of the methods originally proposed


Title: Efficient Homomorphic Encryption for Private Computation in the Cloud. EXTENSION from 2015

Professor and student: Diego de Freitas Aranha; Hilder Vitor Lima Pereira

Summary: The project’s goal is to develop privacy-preserving versions of two well-known machine learning algorithms: the k-NN classifier and the Principal Component Analysis. The new research proposal aims to improve current homomorphic encryption systems in both quantitative and qualitative ways, by improving the performance of constructions and implementations for lattice-based SHE/FHE schemes and extending current constructions to perform better in multi-user settings.


Chile

Title: Engineering Compressed Random Access Memories

Universidad de Chile

Professor and student: Gonzalo Navarro; Joshimar Córdova

Summary: In this project, the team aims to explore efficient algorithms to implement Compressed Random Access Memory (CRAM) computers. With this, data is always maintained and operated upon in compressed form, instead of compressing data in and out of memory, which implies additional processing.


Title: Gradual Security Typing for the Web. EXTENSION from 2015

Universidad de Chile

Professor and student: Éric Tanter; Raimil Cruz

Summary: In the first stage of their investigation, the team developed a gradual security type system to help programmers write applications that avoid privacy leaks of confidential information. For this second year, they plan to extend their implementation to support objects, explore declassification policies and how to integrate them in a gradual security language, as well as develop case studies and experiments


Colombia

Title: Learning Dynamic Action Units for Three-dimensional Facial Expression Recognition. EXTENSION from 2015

Universidad de los Andes

Professor and student: Pablo Arbelaez; Andrés Felipe Romero

Summary: This project seeks to improve human-computer interaction by analyzing automated human facial expression with deep learning techniques and making use of color, depth and temporal information. A system for emotion classification was developed that allows generalizing across camera viewpoints and scales, as well as by subjects’ ethnicity, age and gender. In the second phase of the project, the research will use the general emotion recognizer as high-level representation, and as the basis of a deep learning architecture for action unit detection with a meta-dataset that can merge the three main publicly available databases for emotion recognition


Mexico

Title: Predicting Zika Epidemics Using Social and Vectorial Contact Networks

Instituto Tecnológico y de Estudios Superiores de Monterrey

Professor and student: Edgar Emmanuel Vallejo Clemente; Héctor Manuel Sánchez Castellanos

Summary: The goal of this project is to extend a computational model for predicting Zika infection risk estimates in small populations to include social and epidemiology networks. This will allow the prediction of Zika outbreaks in advance, enabling the timely implementation of disease control interventions.


Title: Interconnected Dual Biosensor for Type II Diabetes Mellitus. EXTENSION from 2015

Universidad Nacional Autónoma de México

Professor and student: Catalina Elizabeth Stern Forgach; Jehú López Aparicio

Summary: This project seeks to develop a dual sensor that diagnoses, prevents and monitors the evolution of Type II Diabetes Mellitus. In the past year, the research team has made enormous progress: first, the associated error that other commercial glucometers have when measuring multiple glucose has been reduced; secondly, the measurement of insulin will be made with high sensitivity and specificity through a novel technique developed this year, that uses saliva instead of blood. The model proposes a new two-part device: a biocompatible disposable platform for the biological fluid, where the detection of glucose and insulin will be carried out, and a second permanent part that will have the electronics and data transmission part


Peru

Title: Facilitate the Diagnostics of Tuberculosis in Low Resources Settings Using Mobile Technologies and Artificial Intelligence

Universidad Peruana Cayetano Heredia (UPCH)

Professor and student: Mirko Zimic; Jorge Coronel

Summary: The researchers propose to harness appropriate technology to allow MODS, one of the very best non-commercial tests for tuberculosis and multidrug resistance, to be used in low-income regions, and to use a 3D printed microscope, artificial intelligence, mobile devices, cloud computing and Internet, to improve test interpretation in remote settings.