Exploring the Relationship Between Vocabulary Scaling and Algorithmic Performance in Text Classification for Large Datasets