脇田 建 / Ken Wakita

脇田 建 / Ken Wakita

数理・計算科学系 情報理工学院 東京工業大学 / School of Computing, Tokyo Institute of Technology

冒険小説の視覚的分析システムの自動合成に向けて「80 日間世界一周」を事例として（第52回 可視化情報シンポジウム＠那覇）

Abstract: This research aims to establish a systematic method for constructing a visual analysis system　for adventure novels. We propose an automated system that generates visual analytics for adventure　novels, adhering to Schneiderman’s mantra. The design supports four abstraction levels: country, chapter,　story, and paragraph. Users can navigate these levels through a zooming metaphor. By hovering over　elements, users can access detailed information about countries, chapters, migrations, characters,　expenditures, remaining days, and story summaries. The system also presents illustrations and character　portraits. We use a large language model (LLM) to analyze the novel and extract relevant information.　This information is then used to generate images and portraits, either directly or as composites.

Keywords: Text analysis, Generative artificial intelligence, Storyline visualization, Visual exploration

Abstract: The purpose of this study is to establish a systematic method of constructing a visual analytics system for adventure travel novels. The system aims to assist in understanding the novel and to encourage communication using this novel as material. For this purpose, we leverage text generation AI, image generation AI, various natural language processing, graph drawing algorithms, optimization, and clustering.

発表資料

デモシステム

デモシステム。200MB くらいのデータがダウンロードされます。このため、デモの読み込みに1分くらいかかります。

関連発表