This study is a successful example ๐ of applying natural language processing and unsupervised learning for materials science trend analysis. We applied the document embedding and density-based clustering to materials science literature. We obtained the comprehensive understanding of scientific topics in materials science without any insertion of expertise. We defined the topic relevance of each paper and identified main topics and academic interests of organisations in a quantitative and time-aware manner. This repository contains the source code and dataset for the following publication:
- ๐ Quantitative Topic Analysis of Materials Science Literature Using Natural Language Processing .
Dataset used in this study is available, and unzip in your own directory.
-
temp_mat_abstract.txt : Sample of the abstract of materials science literature used in this study, which is published between 2017 and 2021.
-
temp_mat_bib.txt : Sample of the bibliographic information of the collected papers, of which columns are doi, title, journal, year.
Our experiment setting is as follows:
-
gensim : 4.1.2
-
spacy : 3.2.4
-
hdbscan : 0.8.28
pip install -r requirements.txt
python code/run.py -dataset <dataset>
Topic Map of Materials Science
National Interests of Materials Science
Text-mined academic interests of Nature Materials, a randomly selected journal
Coming soon!
If you utilise our findings, methods, or results, please consider citing the following paper.
- Choi, J., & Lee, B.* (2024). Quantitative Topic Analysis of Materials Science Literature Using Natural Language Processing Here