Semiotic intelligence and Coronavirus: deep learning and speech analysis in Portuguese-Twitter

• Grant number: 20/15160-7
• Support Opportunities: Scholarships in Brazil - Doctorate
• Start date: April 01, 2021
• End date: March 31, 2026
• Field of knowledge: Linguistics, Literature and Arts - Linguistics
• Principal Investigator: Waldir Beividas
• Grantee: Tulio Ferreira Leite da Silva
• Host Institution: Faculdade de Filosofia, Letras e Ciências Humanas (FFLCH). Universidade de São Paulo (USP). São Paulo , SP, Brazil
• Associated scholarship(s): 23/10101-0 - Hope speech: modeling and validating an automatic text detection system, BE.EP.DR

Abstract

The Coronavirus pandemic has transformed countless scientific terms into the focus of publications for a large part of the Brazilian population. Taken out of their usual contexts, words like COVID, virus, vaccine and Coronavirus have been reframed to express the intimacy of users of the social network Twitter. Between May 2 and September 16, using scrapers, we collected 2.645 million texts published on that social network. However, more than discovering, quantitatively and qualitatively, how the Brazilians related to such terms during the mentioned period, our research intends to develop a semiotic model for the processing of natural language and to deepen the methodology developed during the master's degree. In it, we used Greimasian semiotics and contemporary developments known as 'content prosodization', to develop a method concurrent to sentiment analysis, for the TripAdvisor platform. The project, which unfolds in a specialization in Big Data at the Polytechnic School of USP, awaits a favorable opinion to be patented. Despite this possible first success, the challenges with Twitter are much greater. After all, while evaluation platforms only subsume the stage of / Sanction / (according to the Canonical Narrative Scheme), the microblogging network contemplates all stages (/ Manipulation /, / Competencialization/ and /Performance/, in addition to that one already mentioned). To overcome this challenge, we believe we can use the concept of narrativization of the enunciation or, more precisely, the theme of production. In addition, and from the concepts already mentioned above (and used in the master's degree), we use the idea behind the Portuguese-Twitter term, coined in 2011 as a strategy for the development of a morphosynthetic annotator for the same platform, and the hypothesis of "discursive complexity on the internet", to treat texts produced on Twitter as "another language" (or dialect). With success, our research will be a precursor in offering a methodology to process natural language in digital social networks - which leads us to consider obtaining a new patent. (AU)