What do wordlists tell us? —Adding the dimensions of sense and difficulty levels into the existing English wordlists

本研究主要藉由探討國內與國外現存英文詞表，比較其中之異同，並且從概念化的角度探討基本詞彙的本質。本研究檢視基本詞彙在英文詞網語意網路中的分佈，試圖找出難度和分布之間的關聯。此外，為了將國內常用之英文詞表放入實際與研究上的應用，本研究比較國內外詞表對於全民英檢閱讀測驗五十五篇(初級、中級、中高級)的涵蓋率，並藉此檢視閱讀測驗文章中字彙的難度。 研究結果顯示，總共有1299個詞重複出現在每個詞表中，基本詞彙比進階詞彙有更多的詞義，其中基本動詞的平均詞義高達七個，多於名詞和形容詞(平均各有五個詞義)。至於語意距離方面，基本動詞在語意階層中的上層，顯示基本動詞的詞意較進階動詞廣泛。但名詞卻沒有顯著的模式，可能的原因為名詞若是過於抽象或具體實都容易造成難度提升。計算詞表涵蓋率的部分，所選之國內外詞表對於全民英檢閱度測驗的文章均有約90%的覆蓋率,統計結果也顯示各個詞表的覆蓋率沒有顯著差異。此外，初級的文章中還是出現詞表中偏向進階的單字。最後，研究結果也顯示文章難度越高，則平均的詞義數目越低。本篇研究架設了一個網路介面供使用者計算不同詞表之涵蓋率、字彙分佈以及平均詞義數目，並且根據詞義數目與語意距離，將大考七千單的動詞重新排序，提供一個從新的維度之下產生的動詞詞表。 本研究主要貢獻為加入詞義與語意距離的維度來計算字詞的基本性，學習者與教學者也可利用詞網得到更多基本詞彙之多義性與其語意關聯。本研究也希望能夠幫助教學者與學習者分析不同詞表與文章間的涵蓋率。This study aims to examine the existing English wordlists used in Taiwan and abroad, trying to make comparison between them and exploring the nature of basic words from a conceptual perspective. Basic words are mapped to WordNet semantic hierarchy to see their semantic distance from the root node, and the senses of basic words in the WordNet are also extracted. Furthermore, to put local English wordlists into application, this study examines the text coverage of GEPT wordlist, Vocabulary 7000, BNC lists and NGSL&NAWL on GEPT reading comprehension tests (55 articles), and attempts to examine the word distribution of given articles. The results find that there are 1299 words appearing in every wordlist selected. It is also discovered that basic words tend to have more senses in the English WordNet. The average sense number of basic verbs is up to seven, while basic nouns and adjectives have five senses on average. Furthermore, the distances between basic verbs and their root nodes are lower than those of advanced verbs, and this suggests that basic verbs tend to be more general in meaning. On the contrary, basic nouns do not share the same pattern as basic verbs do; the possible reason may be that when the meanings of nouns become either too general or too specific, they become more unfamiliar to learners. Finally, this study analyzes the coverage of different lists on GEPT reading comprehension tests, examining the word distribution. The coverage of all the lists reaches around 90%, and according to statistical analysis, all the four lists perform quite the same. It is also discovered that the average sense number of elementary level articles is higher than that of intermediate levels. When the difficulty of the articles increases, the average sense number then decreases. Finally, an online interface was built to help users analyze word distribution of different lists. A new verb list is also provided according to the sense numbers and semantic distances. The findings of the study suggest that sense numbers and semantic distance may be taken into consideration in defining word basicness. In addition, learners and teachers should be aware of the number of senses in acquiring basic words, and also the semantic relationship between them by the help of WordNet. Further research may explore deeper into the other semantic relations between basic words