Methodology to build labeled corpora and classification models to assess technological readiness: a case study with defense technologies described by texts in pt-br

The Technology Readiness Levels (TRL) scale is a tool for assessing the maturity of technologies. It has been increasingly used by governments and industries to implement tasks such as technological trends detection. The complexity of the assessment process has demanded automated solutions based on classification models that analyze documents to infer the TRL of the technologies described by the texts. However, the lack of corpora with labeled documents has hindered the development of such solutions. To fill this gap, this study proposes a methodology for building corpora labeled w.r.t. the TRL scale. It was applied to a case study and generated a corpus with 168 documents. This corpus was used to develop 30 classification models that hit 51.72% average F1-score.