Evaluasi Kinerja TextRank dan LexRank Berbasis TF-IDF dan Word2vec untuk Text Summarization

Abstrak

Text summarization is a crucial task in natural language processing, aiming to extract essential information from lengthy texts. The choice of summarization method significantly influences the quality of the generated summary. This study evaluates the performance of the TextRank and LexRank algorithms, both combined with TF-IDF and Word2Vec-based word representation techniques. The IndoSum dataset was used as the benchmark, with preprocessing steps including text cleaning, case folding, tokenization, and vector transformation using Word2Vec and TF-IDF weighting. The ROUGE metric was employed to assess summarization quality. Experimental results indicate that the TextRank algorithm, when integrated with TF-IDF and Word2Vec, achieves higher ROUGE scores compared to LexRank in generating extractive summaries.