Pipelines for Text Classification are sequences of tasks needed to be performed to classify documents. The pre-processing phase of these pipelines involves different ways of manipulating documents for the learning phase. This Master Thesis introduces three new steps into the traditional pre-processing phase: 1) Meta-Features Generation; 2) Sparsification; and 3) Selective Sampling. Our experimental results, based on more than 5.600 measurements, show that our proposal can achieve significant gains in effectiveness when compared to the traditional TF-IDF representation (up to 52%) and word embeddings (up to 46%), at a much lower cost (9.7x faster). Our Master Thesis also includes a thorough and rigorous evaluation of the trade-offs between cost and effectiveness associated with the introduction of these new steps into the pipeline, as well as a comprehensive comparative experimental evaluation of many alternatives. This thesis falls under the topics of (i) Document Management and Classification, (ii) Information Retrieval Models and Techniques, (iii) and Text Database of the SBBD Call for Papers.