1:30 PM - 1:50 PM
[2P4-GS-11-01] Extracting Scholarly Tweets From Researcher Accounts and Classifying Their Areas of Expertise
Keywords:social media, twitter, scholarly data
This study presents an approach to extracting scholarly tweets and classifying their areas of expertise. Our methodology involves training a classifier on a dataset comprising diverse research projects conducted in Japan and tweets from non-researchers. To build our classifier, we compared two pre-trained language models: RoBERTa, which was trained on the CiNii Articles corpus, and BERT, which was trained on the Japanese version of Wikipedia. Our experiments revealed that the former model achieved better classification accuracy than the latter. We also assigned each researcher account with an area of expertise based on the content of their tweets and visualized the results using a social network between researchers.
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.