Abstract:

The United Nations, with the aim of ensuring the sustainable development of the planet, issued a global call for action to end poverty, protect the environment and climate and ensure that people everywhere can enjoy peace and prosperity. In 2015, a set of 17 Sustainable Development Goals (SDGs) was created to be achieved by 2030, which encompasses goals and indicators that represent a set of actions necessary to guarantee the sustainable development of the planet. With the challenge of implementing the SDGs, governments and organizations began to develop actions to implement and monitor these objectives in relation to their fulfillment and evolution. These processes, in addition to the results, generate the need to process a large amount of information, from different sources and formats, including texts, numerical values, tables, etc. Higher Education Institutions (HEIs) are important agents in promoting and supporting the achievement of the SDGs. The literature raises questions about how this contribution can be identified based on data, structured or not, relating to its performance. One of these contributions is present in theses and dissertations resulting from academic research in Postgraduate Studies. To assist in the process of organizing and extracting knowledge from documents, many technological solutions make use of Artificial Intelligence (AI). When working with unstructured data, the use of Natural Language Processing (NLP) and Machine Learning (ML) techniques stands out in many knowledge extraction tasks, especially in the document classification process. In this context, this work aims to propose an analysis framework based on a theses and dissertations classification model using Deep Learning techniques and the use of pre-trained models for Natural Language Processing, such as Bidirectional Encoder Representations for Transformers (BERT). The proposed model was trained with data labeled by SDG experts and successfully used in the process of classifying theses and dissertations of one postgraduate program at the Federal University of Santa Catarina (UFSC), in Brazil, presenting satisfactory performance, allowing association to the corresponding SDGs and extracting data from advisors and students responsible for their development. The results of this model are relevant for mapping an institutional profile in relation to the contribution to the SDGs and are part of a general framework under development to guide and assist this process within HEIs. The next steps of the research involve including the thesis and dissertations from more than one hundred postgraduate programs and modeling the analysis of bibliographic production, research projects and outreach projects for the entire HEI with regard to its contributions to the SDGs.

Keywords:

Sustainable Development Goals, Higher Education Institution, artificial intelligence, Deep Learning, BERT.