Fine-tuning Wav2Vec2 for Classification of Turkish Broadcast News and Advertisement Jingles

No Thumbnail Available

Date

2023

Journal Title

Journal ISSN

Volume Title

Publisher

Institute of Electrical and Electronics Engineers Inc.

Open Access Color

OpenAIRE Downloads

OpenAIRE Views

Research Projects

Organizational Units

Journal Issue

Abstract

The accurate classification of news and commercial jingles is essential for the automated generation of broadcast flow. Currently, in press companies, editors manually label the start and end times of news and advertisements, which incurs both cost and time loss. Although the method of extracting fingerprints of news and commercial jingles has been employed to detect jingles on a channel basis and automatically classify news and commercial music, this approach falls short when it comes to classifying new jingles produced by channels. In this study, we created a new dataset by extracting segments of commercial and news jingles from TV channels in Turkey. We analyzed the most effective second interval for classifying news or commercials, resulting in an impressive accuracy score of 98.18%. By leveraging this dataset and conducting extensive analysis, we have made significant progress in accurately classifying news and commercial jingles. This research can potentially save press companies costs and time by automating the classification process. © 2023 IEEE.

Description

Keywords

advertising, binary classification, jingle, news, speech classification, wav2vec2

Turkish CoHE Thesis Center URL

Fields of Science

Citation

0

WoS Q

N/A

Scopus Q

N/A

Source

2023 Innovations in Intelligent Systems and Applications Conference, ASYU 2023 -- 2023 Innovations in Intelligent Systems and Applications Conference, ASYU 2023 -- 11 October 2023 through 13 October 2023 -- Sivas -- 194153

Volume

Issue

Start Page

End Page