Keyword Extraction Based on Text Hierarchical Structures

• Main Author: RESA SEPTIARI
• Other Authors: Hsing-Kuo Pao
• Format: Others
• Language: en_US
• Published: 2016
• Online Access: http://ndltd.ncl.edu.tw/handle/40398443059460877685

碩士 === 國立臺灣科技大學 === 資訊工程系 === 104 === Keyword extraction is one of the most popular topics in text mining and information retrieval for the purpose of summarization, recommendation, and categorization on texts. Given a text, the goal of keyword extraction is to know which word is important and informative as representatives to describe the text. Keyword extraction has been improved rapidly in recent years as there are a rich set of methods proposed for the task. We propose a novel keyword extraction method of keyword extraction based on machine learning techniques and the consideration of the hierarchical structures of text to further improve the extraction performance from previous methods.We believe that the decision of choosing keywords is based on the focused themes of texts, the authors’ personal preferences, as well as the structures of the texts such as how to present a main topic in each part of texts. We test the proposed method on the data set that consists of a set of research papers for the keyword extraction task by computing the recall, precision, and F-measure in each trial. As a result, the proposed method shows effectiveness in terms of the prediction power and efficiency in terms of the processing time in most cases compared to previous methods.