<![CDATA[The growth of digital health information has led to a significant increase in Indonesian-language health news, creating a need for information retrieval systems that can deliver fast and relevant search results. However, searching Indonesian-language content is challenged by morphological variation caused by affixation, as well as typographical errors (typos), both of which can degrade search quality. This study aims to implement a morphology-based Tolerant Retrieval method combined with the Vector Space Model (VSM) to improve the effectiveness of searching Indonesian health news. The system is built through a preprocessing pipeline consisting of case folding, cleaning, tokenizing, stopword removal, and stemming using the Nazief-Adriani algorithm via the PySastrawi library. Documents are represented using TF-IDF weighting, and similarity is measured using Cosine Similarity. To increase tolerance to word variation and typing errors, the system applies a Bigram Index, a Permuterm Index, and the Damerau-Levenshtein algorithm as mechanisms for query correction and expansion. The dataset consists of 958 Indonesian health news documents and 45 test queries divided into three main categories (normal, morphological variation, and typo), each accompanied by document relevance ground truth. Evaluation results show that the system achieves an average Precision of 85.10%, Recall of 65.11%, and F1-Score of 68.72% at a top-K of 50 documents. These results indicate that integrating morphology-based Tolerant Retrieval with the Vector Space Model improves search result relevance and effectively handles morphological variation and typing errors in Indonesian health news retrieval]]>
Copyrights © 2026
