This paper addresses the challenge of distinguishing header metadata from body content in Persian elec- tronic theses and dissertations. Accurate classification of these sections aids tasks such as metadata extraction from scientific documents and plays a crucial role in in- creasing the efficiency and retrieval of information in digital libraries. Several machine learning models were employed to achieve this goal. Additionally, five distinct feature types were utilized: Heuristic, Sequential, Lexical, Formatting, and Geometric. The dataset consisted of nearly 230,000 paragraphs extracted from 106 Persian ETDs, with the metadata class representing only 8.6%. After preprocessing, Random Forest slightly outperformed SVM and Naïve Bayes. Moreover, our findings indicate that features of sequential type notably impact the classification metrics.

Keywords