「適性文本閱讀支援系統」之建置與應用研究-「適性文本閱讀支援系統」之建置與應用研究_團隊計畫
No Thumbnail Available
Date
2014-07-31
Authors
宋曜廷
陳柏琳
柯華葳
張道行
蘇宜芬
Journal Title
Journal ISSN
Volume Title
Publisher
Abstract
本研究目的在建置一中文「適性文本閱讀支援系統」(Readability Evaluation and Analysis of Chinese Text, REACT)，本系統包括三個子系統，分別是「文本可讀性計量子系統」、「專家文本可讀性分級子系統」以及「讀者閱讀年級適性評量子系統」。在「文本可讀性計量子系統」中，我們將透過潛藏語意分析技術（latent semantic analysis, LSA）發展語意特徵，以自然語言處理與資訊處理技術發展中文句法剖析器發展語法特徵，並結合資訊工程與計算語言學理論，建置語言模型進行文本可讀性計量。各類可讀性特徵產出後，可讀性指標自動分析器可自動擷取數值經人工智慧算則建立可讀性數學模型，對文本年級提供標記。在「專家文本可讀性分級子系統」我們納入與閱讀理解有關的可讀性變項，並據此編製專家問卷，進行專家文本分級。「讀者閱讀年級適性評量子系統」則可評量讀者閱讀水準，此子系統提供子計畫五效度驗證時的讀者閱讀能力的評量，以提供「文本可讀性計量子系統」與「專家文本可讀性分級子系統」對文本年級預估的準確性之回饋。本研究將透過總計畫與五個子計畫完成。總計畫負責協調各子計畫研究進程、系統整合與調整，子計畫一以LSA 技術發展語意特徵，子計畫二以句法剖析器發展語法特徵，子計畫三則以語言模型進行文本可讀性推估，子計畫四發展「讀者閱讀年級適性評量子系統」，子計畫五發展「專家文本可讀性分級子系統」，並以眼球追蹤儀進行效度驗證。此外，本研究將持續擴充涵蓋各類領域知識的文章資料庫，並根據讀者反應調整系統，提供研究者、教師、家長與讀者適性文本需求與配對建議，建構全球最有競爭力的中文適性閱讀平台。
The purpose of this study is to construct the ’Readability Evaluation and Analysis of Chinese Text’, REACT. This system is composed of three subsystems as follows. The first subsystem, Text Readability Measuring System, can analyze and produce the text-based features, and label the text grade level. The whole subsystem is composed of a mathematical model and the core measuring tool, called Chinese Readability Index Explore, in which Latent Semantic Analysis is responsible for the extraction of semantic information and the Chinese syntactic parser, developed with the techniques of natural language processing and information engineering, can produce the syntactic features. The language model, which is constructed based on the technologies of information engineering and the theories of computational linguistics, integrates the various text-based features and estimates the grade level of texts. The second subsystem, Text Leveling System, designs a structure, based on the variables relevant to reader-orientation and comprehension. With this structure, this system can produce the grade level of texts as well. The third subsystem, Adaptive Reading Evaluation System works on the evaluation of readers’ reading levels. The outcomes of this subsystem can be further combined with psychological experiments and used as an accuracy checker to compare the estimations of text level produced by the first and second subsystems.
The purpose of this study is to construct the ’Readability Evaluation and Analysis of Chinese Text’, REACT. This system is composed of three subsystems as follows. The first subsystem, Text Readability Measuring System, can analyze and produce the text-based features, and label the text grade level. The whole subsystem is composed of a mathematical model and the core measuring tool, called Chinese Readability Index Explore, in which Latent Semantic Analysis is responsible for the extraction of semantic information and the Chinese syntactic parser, developed with the techniques of natural language processing and information engineering, can produce the syntactic features. The language model, which is constructed based on the technologies of information engineering and the theories of computational linguistics, integrates the various text-based features and estimates the grade level of texts. The second subsystem, Text Leveling System, designs a structure, based on the variables relevant to reader-orientation and comprehension. With this structure, this system can produce the grade level of texts as well. The third subsystem, Adaptive Reading Evaluation System works on the evaluation of readers’ reading levels. The outcomes of this subsystem can be further combined with psychological experiments and used as an accuracy checker to compare the estimations of text level produced by the first and second subsystems.