- 期刊
AI人工知能技術を援用した文学研究の試み－ストーリー展開特徴抽出による小説の可視化－
AI人工智慧技術在文學研究中的應用－基於故事發展特徵的小說可視化分析－｜The Application of Artificial Intelligence Technology in Literary Research: A Visualization Analysis of Novels Based on Story Development Features
摘要
近年、General-purpose computing on graphics processing units（GPGPU）技術の発展を含むコンピューターの性能向上に伴い、深層学習（Deep Learning）に代表される機械学習（Machine Learning）の性能が大幅に向上し、そして、文学研究などの芸術の分野への応用も広がりを見せている。単独作品の精緻な検討を行うという従来の文学研究スタイルとは異なる、巨大コーパスを扱い文学空間を数理的に扱うDigital Literature（DL）の研究の多くは語彙の特徴や品詞の分布といった単語ベースの手法であり、文を基本単位とした手法は少ない。しかし、語彙の特徴のみを検証しても物語の表層的な特徴しか見られないため、文の意味のような深層的な特徴にも目を向ける必要がある。そこで、本研究は小説の展開に対する読解をさらに掘り下げる試みとして、文を基本単位とし、日本語評価極性辞書を用いて文のネガポジ判定を行った。その結果、「緑色の獣」はネガティブ性が強く、一方、「氷男」はポジティブ性が強い。このことにより、この二つの小説が類似していない結果が得られた。感情表現が表れている箇所が視覚化されて明確になり、感情の変遷を客観的かつ定量的に示して検証することができたことから、提案手法の有用性を確認できた。
並列摘要
近年來圖形處理器（GPGPU）運算技術的改良帶來電腦效能提升，以電腦輔助進行研究被大量應用人文學研究等相關領域。不同於傳統文學研究著重於研究者主觀分析居多，數位文學（Digital Literature，DL）提供了定量、客觀的文學研究例證。過往的數位文學研究多以語詞為主，強調詞頻分析，並未考慮小說內的情緒流動。因此本論文嘗試以村上春樹的作品為分析文本，透過情感分析的技術進行小說可視化分析。通過對村上春樹作品〈綠色之獸〉、〈冰男〉情感變化的分析，得出了〈綠色之獸〉整體情感趨勢以負面為主，而〈冰男〉則以正面為主，兩部小說的情感流向有很大的差異性之結論。由此可知，透過情感分析，可使小說文字間複雜的情感流動更為清晰。本研究作為一種新的嘗試，驗證了情感分析技術對文學研究和文學作品分析的可行性。｜Recent improvements in GPGPU computing technology have led to increased computer performance, and computer-assisted research has been used in a wide range of fields, including humanities research. Unlike traditional literary research that focuses on subjective analysis, Digital Literature (DL) provides quantitative and objective examples of literary research. In the past, digital literature studies were mostly language-oriented, emphasizing the analysis of word frequency, without considering the emotional flow in the novel. Therefore, this paper attempts to analyze Haruki Murakamis works as the text and visualize the novel through the technique of sentiment analysis. The analysis is carried out through the technique of sentiment analysis. Through the analysis of the emotional changes in Haruki Murakamis works "The Little Green Monster" and "The Ice Man", it is concluded that the overall emotional trend of "The Little Green Monster" is mainly negative, while "The Ice Man" is mainly positive, and there is a great difference in the emotional flow of the two novels. Through emotional analysis, the complex emotional flow between the novels can be made clearer. This paper is a new attempt to verify the feasibility of sentiment analysis techniques for literary research and analysis of literary works.｜Recent improvements in GPGPU computing technology have led to increased computer performance, and computer-assisted research has been used in a wide range of fields, including humanities research. Unlike traditional literary research that focuses on subjective analysis, Digital Literature (DL) provides quantitative and objective examples of literary research. In the past, digital literature studies were mostly language-oriented, emphasizing the analysis of word frequency, without considering the emotional flow in the novel. Therefore, this paper attempts to analyze Haruki Murakamis works as the text and visualize the novel through the technique of sentiment analysis. The analysis is carried out through the technique of sentiment analysis. Through the analysis of the emotional changes in Haruki Murakamis works "The Little Green Monster" and "The Ice Man", it is concluded that the overall emotional trend of "The Little Green Monster" is mainly negative, while "The Ice Man" is mainly positive, and there is a great difference in the emotional flow of the two novels. Through emotional analysis, the complex emotional flow between the novels can be made clearer. This paper is a new attempt to verify the feasibility of sentiment analysis techniques for literary research and analysis of literary works.