Corpus-based Sentence Deletion and Split Decisions for Spanish Text Simplification

This study addresses the automatic of texts in Spanish in order to make them accessible to people with cognitive disabilities. corpus analysis of original and manually simplified articles was undertaken in order to identify quantify relevant operations to be implemented a text simplification system. The articles were compared at sentence and text level by of automatic feature extraction and various learning classification algorithms, using three groups of features (POS frequencies, syntactic, and text complexity measures) with the of identifying features that help separate original from their simple equivalents. Finally, it investigated whether these features can be used decide upon simplification operations to be carried at the sentence level (split, delete, and reduce). classification of original sentences into those be kept and those to be eliminated outperformed the that was previously conducted on the same. Kept sentences were further classified into those be split or significantly reduced in length and those be left largely unchanged, with the overall F-measure to 0. 92. Both experiments were conducted and on two different sets of features: all features the best subset returned by an attribute selection.