This paper introduces TextRank, a graph-based ranking model for text processing, and demonstrates its effectiveness in natural language applications. The authors propose two innovative unsupervised methods for keyword and sentence extraction, showing that TextRank's results are competitive with state-of-the-art systems on established benchmarks. TextRank is based on the concept of "voting" or "recommendation," where the importance of a vertex (text unit) is determined by the votes cast for it and the importance of the voters. The model is applied to two tasks: keyword extraction, which involves selecting keyphrases representative of a text, and sentence extraction, which aims to identify the most important sentences for extractive summarization. The paper evaluates TextRank using a dataset of 500 abstracts from the Inspec database and 567 news articles from the Document Understanding Evaluations 2002 (DUC). Results show that TextRank achieves high precision and F-measure in keyword extraction and outperforms other systems in sentence extraction tasks. The authors attribute TextRank's success to its ability to consider global information from the entire text, making it highly portable to various domains, genres, and languages.This paper introduces TextRank, a graph-based ranking model for text processing, and demonstrates its effectiveness in natural language applications. The authors propose two innovative unsupervised methods for keyword and sentence extraction, showing that TextRank's results are competitive with state-of-the-art systems on established benchmarks. TextRank is based on the concept of "voting" or "recommendation," where the importance of a vertex (text unit) is determined by the votes cast for it and the importance of the voters. The model is applied to two tasks: keyword extraction, which involves selecting keyphrases representative of a text, and sentence extraction, which aims to identify the most important sentences for extractive summarization. The paper evaluates TextRank using a dataset of 500 abstracts from the Inspec database and 567 news articles from the Document Understanding Evaluations 2002 (DUC). Results show that TextRank achieves high precision and F-measure in keyword extraction and outperforms other systems in sentence extraction tasks. The authors attribute TextRank's success to its ability to consider global information from the entire text, making it highly portable to various domains, genres, and languages.