This work proposes a recommendation system for official diaries content available in “Querido Diário”.
For tokenization of the input data, we use the tokenizer of the trained agent in BERTimbau (BERT agent trained in Portuguese).
We use BERTimbau's embeddings to feed a KNN classifier, which is used in the semantic content recommendation task.