Visualization of Audio Transcription as High-low Graphs

Abstract

The purpose of this paper is to propose a method which visualizes the contents of audio transcription, such as interviews, podcasts and chats as High-low Graphs. The method extracts important words from audio transcription using morphological analysis and TF-IDF. Wikipedia is used to generate proper name dictionary for morphological analysis and used to count frequency of TF-IDF. High-low graphs of important words are made by Kernel density estimation, which are enable to show the feature of the audio transcription.