拓樸語言學取向的概念分析:以社區一詞為例

Abstract

在2003年，台灣社會受到SARS傳染病的嚴重衝擊。從SARS的事件中，可以觀察到在政令下達的過程中沒有人能忽略跨領域合作的重要性。然而，合作需要溝通。特別是當一個團隊擁有不同的專業者；在許多跨領域的問題情境中，他們必須重視對於某些事物的共同理解。這就是為什麼本研究要針對這個事件且以「社區」一詞為例，提出一個半自動化的方法去把隱含在脈絡當中的概念結構視覺化。 為了分析「社區」一詞，這個研究以運算語言學與建構本體知識工程為基礎提出一個分析性的機制。所謂知識本體的概念結構不外乎區辨概念以及概念與概念之間的關聯性。因此，在不同知識領域(即不同類型的說話者)之間要區辨概念的方法程序，本研究提出的步驟規劃如下： 1.蒐集資料: 在SARS期間的新聞報導與相關文章收集起來，檢索出所有包含「社區」一詞的語句作為基本的分析單元。 2.進行分析: 這個機制為了增加其資料分析的可操作性用來帶動語句型式的轉換。然後，把一個語句分解成tuple來掌握概念與概念之間之運作關係。 3.描繪概念輪廓: 本研究建構了一個基模用來描繪每個語句的概念輪廓。這個基模是一個架構不只可以把一連串語彙的概念予以結構化，而且盡可能地把語境脈絡內隱的詮釋給外顯化。 4.繪製概念圖: 本研究根據分析最後會產生一個形式良好的資料表單來體現其既有文本中的語境脈絡。研究的結果會就語詞外延的意義歸納分類。而且，本研究運用了「GRAPHVIZ」的圖形工具語言描繪其星狀拓樸圖。 本研究的結果是以這些星狀拓樸圖來觀察不同領域之間對於「社區」一詞的不同領域知識本體的轉移現象。為了使研究的方法與結果具有可信度，本研究採取了兩個策略；針對分析結果，本研究透過雙盲測試來檢核並保留可以採信的部分。針對分析方法的合理性，本研究透過全球語料規模最大的「中文詞彙特性速描系統」(CSE)所查詢針對「社區」一詞所搭配的詞組作為基準線來進行對照。還有就是應用一些統計數據的指標來觀測其變異性的程度。然後，運用質性變異指標(IQV)與卡方檢定來評估在變異性的向度上不同領域之間的表現。

In 2003, Taiwan came under the attack of Severe Acute Respiratory Syndrome (SARS). The SARS epidemic had a huge impact on Taiwan society. From the event, we observed that no one can be regardless of interdisciplinary collaboration in the delivering of public policies. However, collaboration needs communication. Especially when a team comprises various professionals; they have to value joint comprehension under trans-disciplinary situations. That is the reason for the research to propose a new semi-automatic skill to visualize implicated contexts in case of the term 「社區[shèqū]」(“community”). To analyze “community”, this research proposes an analytical mechanism based on computational linguistics and ontological engineering. Hence, the way to identify concepts among knowledge domains (various types of speakers) is a step-by-step procedure: 1.Collecting Data: after retrieving all documents, we keep all the sentences containing the term “community”. A sentence is an entry to be analyzed. 2.Implementing analysis: for data processing, the mechanism is made for carrying transformations to increase the maneuverability. Then, a sentence would be decomposed as tuples to reveal their concepts and operators. 3.Profiling Concepts: the corpus is analyzed to identify concepts and their relationships. Thus, the research creates a schema for profiling each sentence. The schema is a framework not only to structure our knowledge in lexical chains, but also to explicate their interpretations in context. 4.Mapping conceptual graphs: the study generates a well-formed database to exhibit the given texts (community) within their contexts according to the analysis. Results are categorized as the extended senses. Furthermore, the research applies the graphic tool GRAPHVIZ which conveys the star topologies. Conclusively, all the star topologies show various kinds of conceptual structures of “community”. Apparently, there seems to be the shifting of domain ontologies while contrasting to knowledge domains and relationships. To be convinced, we apply two strategies to evaluate the results and the method of the research. For the result, the research uses double-blind test to check the transformation from sentence to tuple. For the method, the research makes a contrast to the Chinese Sketch Engine (CSE) which owns the largest readership globally. Also, we apply several statistic indexes to measure the difference between CSE and the four domains. Then, by ways of Index of Qualitative Variance (IQV) and Chi-square test, the research has verified its method on the dimension of variance and diversity.