GENERATING SCENE DESCRIPTOR FROM INDONESIAN NARRATIVE TEXT

• Main Author: Aufa Helfiandri, Muhammad
• Format: Final Project
• Language: Indonesia
• Subjects: Teknik (Rekayasa, enjinering dan kegiatan berkaitan)
• Online Access: https://digilib.itb.ac.id/gdl/view/47949
• Institution: Institut Teknologi Bandung

Animation is one of media that are frequently used in the entertainment world. Production of animation starts from scripts which then used to create storyboards which in turn used to create animation scenes. This research aims to find an automatic way to generate animation scenes from text input. No existing research has been conducted for generating animation automatically from Indonesian narrative text. Our system consists of three parts, namely: Preprocessing, Narrative Understanding, and Visualization. However, this system focused only on Preprocessing and Narrative Understanding. This system employs Constituency Parsing as the main Parsing technique, with supporting techniques such as Dependency Parsing and Word Embedding. Testing was conducted to the 420 narrative text dataset gained from Wattpad for Indonesian. The performance results show that this system is capable of generating scene description from narrative text with F1 scores for object, object count, pose, relation and background are 0.9541, 0.8294, 0.8294, 0.5216 and 0.3486