Research on Bert-based topic generator

Yuanyuan Cai; Tongxin Li; Qingchuan Zhang

doi:10.1117/12.2639761

19 October 2022 Research on Bert-based topic generator

Yuanyuan Cai, Tongxin Li, Qingchuan Zhang

Proceedings Volume 12294, 7th International Symposium on Advances in Electrical, Electronics, and Computer Engineering; 122943T (2022) https://doi.org/10.1117/12.2639761
Event: 7th International Symposium on Advances in Electrical, Electronics and Computer Engineering (ISAEECE 2022), 2022, Xishuangbanna, China

Abstract

In order to quickly and accurately identify the frontier research hot spots in the subject field and provide accurateprediction service for scientific researchers, this paper takes the scientific and technological literature in the fieldof coldchain logistics as the research object, and proposes a BTG (Bert-Based Topic Generator) model, which extracts thesubject words from the titles, keywords and abstracts of the literature. In this model, the pre-trained language model Bert is used to obtain the word embedding, which is dimensionally reduced by the UMAP algorithm, HDBSCANclusteringalgorithm is used for clustering, and TF-IDF weight is calculated into the core subject words in each clustering category. This study conducts comparative experiments on HDBSCAN, K-means and DBSCAN clustering algorithms byusingevaluation criteriones such as degree of cohesion and degree of separation. According to the experimental results, theDBSCAN algorithm outperform other clustering algorithms based on Separation criterion, Intra-Cluster part ofDispersion, Inter-Cluster part of Dispersion and PR criterion.

Citation Download Citation

Yuanyuan Cai, Tongxin Li, and Qingchuan Zhang "Research on Bert-based topic generator", Proc. SPIE 12294, 7th International Symposium on Advances in Electrical, Electronics, and Computer Engineering, 122943T (19 October 2022); https://doi.org/10.1117/12.2639761

ACCESS THE FULL ARTICLE

INSTITUTIONAL
Select your institution to access the SPIE Digital Library.

SELECT YOUR INSTITUTION

PERSONAL
Sign in with your SPIE account to access your personal subscriptions or to use specific features such as save to my library, sign up for alerts, save searches, etc.

PERSONAL SIGN IN

No SPIE Account? Create one

PURCHASE THIS CONTENT

SUBSCRIBE TO DIGITAL LIBRARY

50 downloads per 1-year subscription

Members: $195

Non-members: $335 ADD TO CART

25 downloads per 1 - year subscription

Members: $145

Non-members: $250 ADD TO CART

PURCHASE SINGLE ARTICLE

Includes PDF, HTML & Video, when available