BERTopic
Grootendorst (2022) ๊ฐ ์ ์ํ embedding-based topic modeling ๊ธฐ๋ฒ. 4-step ํ์ดํ๋ผ์ธ: (1) Sentence-BERT (Devlin et al. 2019) embedding ์ผ๋ก ๋ฌธ์๋ฅผ dense vector ๋ก ๋ณํ, (2) UMAP ์ผ๋ก ์ฐจ์ ์ถ์, (3) HDBSCAN ์ผ๋ก density ๊ธฐ๋ฐ cluster ์์ฑ, (4) c-TF-IDF ๋ก cluster representative term ์ถ์ถ. LDA ยท DTM ๋ฑ frequency-based (Bag-of-Words) ๋ฐฉ๋ฒ๊ณผ ๋ฌ๋ฆฌ contextual semantic ๋ณด์กด. Identifying Interdisciplinary Emergence in the Science of Science: Combination of Network Analysis and BERTopic ์์ ํ์ ๊ฐ ๊ณผํ ๋ถ์ ๋ถ์์ ์ฌ์ฉ.
์ธ์ ๊ทธ๋ํ
1-hop ์ด์ 4๊ฐ
- ์ธ๋ฌผ 1
- ๋ ผ๋ฌธ 3
ํ = ํ๋/์ถ์ ยท ๋๋๊ทธ = ์ด๋ ยท hover = ๋ผ๋ฒจ ยท ํด๋ฆญ = ํ์ด์ง ์ด๋
์ด ๋ฌธ์๋ฅผ ๊ฐ๋ฆฌํค๋ ํ์ด์ง
๋ ผ๋ฌธ (3)
- Disruptive technologies for knowledge management: bibliometric review and patent analysis
- Exploring knowledge management technologies to enhance sustainability and mitigate technostress from a collaborative perspective
- Identifying Interdisciplinary Emergence in the Science of Science: Combination of Network Analysis and BERTopic