JSAI2024

Presentation information

Organized Session

Organized Session » OS-5

[2T6-OS-5c] OS-5

Wed. May 29, 2024 5:30 PM - 6:30 PM Room T (Room 62)

オーガナイザ:荒井 ひろみ(理研AIP)、小山 聡(名市大)、鹿島 久嗣(京大)、堤 瑛美子(東大)、森 純一郎(東大)

5:50 PM - 6:10 PM

[2T6-OS-5c-02] An Extension of Dawid-Skene Model Considering Diversity of Labels

〇Kyohei Atarashi1, Hiromi Arai1, Satoshi Oyama2,1, Daisuke Hatano1 (1. RIKEN Center for Advanced Intelligence Project, 2. Nagoya City University)

Keywords:Machine Learning, Crowdsourcing, Fairness

Crowdsourcing is a popular way to obtain labeled data at relatively low cost. Existing methods for inferring true labels from crowdsourced noisy annotations assume that there is one true label for each item, but sometimes it is more appropriate to assume that each item has multiple true labels depending on the attributes of the workers. In this study, we propose a model that estimates the abilities and labels based on worker attributes, which considers a diversity of labels. The proposed model is an extension of the Dawid-Skene model and assumes that there is a true label for each combination of attributes. Experiments on synthesis data, where each item had multiple true labels depending on the attributes of the workers, showed that the existing model underestimated the ability of a minority group, but the proposed method accurately estimated it.

Authentication for paper PDF access

A password is required to view paper PDFs. If you are a registered participant, please log on the site from Participant Log In.
You could view the PDF with entering the PDF viewing password bellow.

Password