The paper “Tagsets and Datasets: Some Experiments Based on Portuguese Language”, written by professor Alexandre Rademaker, from FGV’s School of Applied Mathematics (EMAp), Claudia Freitas (PUC-Rio), Luiza F. Trugo (PUC-Rio), Fabricio Chalub (IBM Research) and Guilherme Paulino-Passos (IBM Research and EMAp alum), received the best paper award at the International Conference on Computational Processing of the Portuguese Language (PROPOR 2018), held from September 24 to 26, in Rio Grande do Sul.

The study discusses the impact of different types of morphosyntactic markings of words on automatic text processing (NLP, natural language processing by computers). It is a multidisciplinary, linguistic and computational study using collections of texts. The work is part of the Portuguese treebank expansion project included in the Universal Dependencies (UD) initiative. Treebanks are collections of texts separated into sentences with morphosyntactic annotations used for training word-processing tools.

Universal Dependencies brings together more than 200 researchers from all over the world to build more than 100 treebanks in 60 different languages. It is a massive effort to build data consistently across different languages for the advancement of NLP. The project includes researchers from Stanford, Google, IBM, as well as major companies and universities around the world.