A discourse analysis framework for civil action decision-making

Abstract

Studying the implications of people’s opinions on social networks has increased the interest of various stakeholders such as the government, leaders, researchers, and citizens. Consequently, human-computer interaction (HCI) has a vital role through civil action to interact with computational models needed to meet these new demands. By conducting several experiments with a corpus of text data collected from Twitter, we plan to create language representation models based on word embeddings to determine the relevance of discourse concerning a topic and detect abrupt changes over time. Thus, for example, citizens could have quantitative information on the relevance of a political leader’s discourse on social issues such as corruption, health, or employment in an electoral process.  Alternatively, in a crisis, the authorities could make decisions on the needs of the people by detecting needs expressed in the context changes of the discourse over time.