- 學位論文
發展知識本體機制於中文新聞解析與分類
Developing Ontological Mechanisms for Chinese News Analysis and Classification
摘要
本研究發展以知識本體為核心進行新聞語意解析與情境分類，即知識架構中納入情境解析機制，使同一則新聞處於不同情境時，產生不同的解讀以利資訊分類。目前常見的文件解析方式，仍侷限於人類對詞彙的認知、詞性判斷以及同義詞的對應等，但對於文件內容更深層涵義(implications)之瞭解仍顯不足。現行的關鍵詞庫僅於個別語彙意義的制式化解釋，但無助於整體文件語意的解讀，文件解析能力應能考量需求者所處情境。故本研究於知識擷取時，需先蒐集相關元素及屬性或已存在的分類架構，利用正規概念分析法(Formal Concept Analysis, FCA)，經由專家分析出元素及屬性之間之顯性及隱性關係，並輔以概念圖整合所有元素、屬性及分類架構間之關聯性。為建構新聞於不同情境時分類的法則，分別依過濾解析階段與情境分類階段兩階段，以轉換資料層級為語意層級，前者採用資源描述架構(Resource Description Framework, RDF)，其目的為改善現有表達詞彙語意的方式；後者採用網路本體語言(Web Ontology Language, OWL)，能加入描述邏輯以協助表達不同情境的知識，使解析新聞詞彙的資料格式由單純的資料層級，提升為具知識表達能力的語意層級。本研究最後以電子產業的「IC零組件」為例，擷取人類專家對於主要生產IC零組件製造商於不同情境下的知識，發展以本體機制解析中文新聞標題，並建置後續不同情境影響的語意分類，作為實證應用。
並列摘要
The common ways of context analysis have been limited to human understanding of vocabularies, speech judgment and synonym mapping, resulting in a lack of understanding of the deeper implications of the content of the text. Based on an ontology knowledge classification structure, our research aims to analyze news semantics and classify news scenarios. The integration of a scenario analysis mechanism into the knowledge structure would allow for different readings of news under different scenarios, benefiting classification of information. In this research we collected relevant knowledge element properties, attributes and any existing classifying structures first. Following Formal Concept Analysis (FCA), we then integrate the elements and dominant/recessive attributes analyzed by the experts into a concept plan which shows the relationship among all the elements, their properties, and classification structures. To enhance the analysis of news contents from an information level to a semantic level, this research utilizes a two-step process, Resource Description Framework (RDF) and Web Ontology Language (OWL); the former improves the expression of vocabularies and the latter adds descriptive logic to help express knowledge under different scenarios. We used the “IC Components” of the electronics industry as a case study to collect the knowledge the experts have regarding the different scenarios the manufacturers encounter. The knowledge was then used to analyze the Chinese news headlines based on the mechanism of ontology and establish a semantics classification as affected by different scenarios afterwards, which will be used as empirical application.