University of Twente Student Theses

Login

Audio-based stylistic characteristics of Podcasts for search and recommendation : a user and computational analysis

Martikainen, Katariina (2020) Audio-based stylistic characteristics of Podcasts for search and recommendation : a user and computational analysis.

[img] PDF
26MB
Abstract:Even though many search engines already excel when it comes to text document retrieval, retrieving spoken content is an entirely different matter. Furthermore, the paralinguistic dimensions of language have been widely neglected in spoken content retrieval. The focus has been only on the content of what is said, ignoring the ways of how things are said. This master thesis argues that also the style of spoken content is important to listeners and that implementing stylistic search and recommendation capabilities could substantially benefit search and recommendation systems of spoken content. This research focuses on searching for and recommending podcasts. We present our results on what kind of stylistic features of podcast content listeners care about, and how these features can form higher-level stylistic categories. Furthermore, we report the results of our experimentation on the technical suitability for using the stylistic features as a basis for automatic podcast recommendation. Our core findings are that podcast listeners have clear ideas on what kind of stylistic content matters to them, and that many of these stylistic features can be used for automatic podcast recommendation based on the information captured from the podcast audio signal.
Item Type:Essay (Master)
Faculty:EEMCS: Electrical Engineering, Mathematics and Computer Science
Subject:02 science and culture in general, 50 technical science in general, 54 computer science
Programme:Interaction Technology MSc (60030)
Link to this item:https://purl.utwente.nl/essays/82810
Export this item as:BibTeX
EndNote
HTML Citation
Reference Manager

 

Repository Staff Only: item control page