文學語意之發掘

Abstract

本文取材自國科會計畫NSC 93-2213-E-002-081，題目與解答為計畫主持人所授，程式來自計畫，其中部分程式經學生修改。莎士比亞文集資料庫為學生整理完成，相關網站部分為學生完成。 我們每天都在面對文句，也或許每天都在創造文句，但當我們仔細去分析，會發現擷取語意的過程本身是很朦朧的。把一個句子裡的字做些不同組合，可能產生的意義就變了；同樣的字放在不同的句子裡，也可能產生不同的結果。所有的技術無論多好，如果只接觸到一些帳面上的統計結果，就無法表達或深入到人類智能的活動。我們延續了實驗室在語意處理上的研究，從時間與序列的關係來處理語意編碼，並且在這些編碼上發展出語意測量的技術，比如語意檢索、語意分類等。 當一套編碼夠好，就會類似一個作家思維的複製。美學是主觀的，所以電腦科學很難跨進其中。但在此我們用客觀的技術去還原一個作家，再讓作家產生主觀的判斷。論文的最後，我們產生了不同作家的複製並展示了他們所產生的不同反應。從這個方向繼續走下去，本編碼方式的無窮的潛力和價值展露無遺。

This work was supported by National Science Council, ROC, under contract number NSC 93-2213-E-002-081. Everyday we face sentences, maybe with creating sentences. However, the acquirement of linguistic is difficult to understand if we want to illustrate the processes and list the rules. Sentences with similar words but dissimilar arrangements may vary on meaning; a word in different sentences may reveal different meaning. No matter how surprising a study results, if the technique only touches statistical analyses, it will be limited for its distance from human intelligent behavior. We continue the research of our laboratory about temporal and sequential semantic encoding, which discover the semantics from a way similar to human reading operation. We also create a new structure of code to approach more the human intelligence, about multi-semantic acquirement. On these semantic codes, we develop and practice many behaviors like semantic comparison, semantic search, and semantic categorization. With a corpus from some writer, when the semantic codes are precise enough, it becomes a clone of this writer's aesthetic. At the last of this paper, we display how different literature masters value one another. The essence of aesthetic is subjective, but we can take advantage of the objective computer science on duplicating one's semantic world and recover his aesthetic. From this way, our research reveals the potential and worth for future studies.