Progressive Feature Engineering for Genre Classification of Short Poems

Textual data may take on a number of formats and presentation styles and serve as an asset for an extensive variety of applications. It is important to thoroughly understand the text’s stylistic cues and other distinguishing features before classifying the documents according to genre, author or emotion. English poetry is a well-known art form that is always expanding and transforming online. As a result, it is critical to classify poetry by genre to enhance poetry recommendation algorithms to provide user-specific relevant suggestions, in E-Commerce, for literary analysis and easy evaluation in competitions. However, determining the genre, particularly for shorter poems, requires more thought and planning. To investigate the task-related skills of eleven contemporary machine learning (ML) models, a feature engineering method coupled with a hybrid dimensionality reduction technique is incorporated. Aggregation of multiple features is performed. The consistent performance of Support Vector Classifier (SVC) and Logistic Regression (LR) in terms of balanced accuracy (ACC), and weighted F-Score ( \(F_1\) ) across the feature groups is notable with LR scoring a mean \(F_1\) of 72.97% and ACC of 72.61%