Motor multifunções: pesquisa terminológica bilíngue e assistente de escrita académica com base em dados científicos abertos

Abstract

In this paper, we will give an overview of the creation process of a multi-function engine that is being developed within the PortLinguE research project (ref. PTDC/LLT-LIG/31113/2017) and reuses scientific data available in open access regime. We will describe the general architecture of the engine, which is based on a Django framework, and the logical model of the engine that will work with BERT machine learning models, as it enables searches that consider context and semantic similarities. The engine has two main functions that are presented in detail: (1) the bilingual terminology search function, capable of identifying translation equivalents of comparable texts taken from scientific repositories (useful to translators and researchers working with specialized languages) and (2) the academic writing assistant function, which relies on the constitution of a phrase bank for European academic Portuguese, through the collection, annotation and analysis of scientific articles taken from national repositories (useful to students seeking to improve their writing in academic contexts).