Related Experiment Video
Updated: Feb 3, 2026

Cloud-Based Phrase Mining and Analysis of User-Defined Phrase-Category Association in Biomedical Publications
Published on: February 23, 2019
A method of inferring the relationship between Biomedical entities through correlation analysis on text
Hye-Jeong Song1,2, Byeong-Hun Yoon1,2, Young-Shin Youn1,2
1School of Software, Hallym University, Chuncheon, South Korea.
Word embedding effectively identifies correlations between diseases, biomarkers, and microorganisms in biomedical texts. This approach accurately predicts research trends, highlighting potential areas for new scientific discovery and investigation.
Area of Science:
- Biomedical Informatics
- Natural Language Processing
- Machine Learning
Background:
- Traditional one-hot word representations in machine learning are high-dimensional and assume feature independence.
- Word embedding excels at capturing semantic meaning and estimating word similarity, crucial for analyzing complex biomedical relationships.
- Biomedical text analysis requires robust methods to uncover hidden correlations between entities like diseases, biomarkers, and microorganisms.
Purpose of the Study:
- To leverage word embedding's similarity estimation capabilities to explore correlations within biomedical texts.
- To identify novel biomarkers and microorganisms associated with specific diseases using advanced natural language processing techniques.
- To analyze relationships between diseases-markers and diseases-microorganisms in biomedical literature.
Main Methods:
- Constructed a biomedical corpus by extracting titles and abstracts from PubMed.
- Applied word embedding, specifically using Canonical Correlation Analysis (CCA) for dimensionality reduction, to represent diseases, markers, and microorganisms.
- Quantified the similarity between disease-marker and disease-microorganism pairs to infer correlations.
Main Results:
- Validated word embedding-derived correlations against Google Scholar search results.
- 85% of top-correlated disease-marker pairs showed significant existing research, confirming the method's predictive power.
- Low-correlation pairs rarely had supporting studies, indicating their novelty or lack of established relationships.
Conclusions:
- Word embedding-based correlation analysis accurately reflects current biomedical research trends.
- High correlations suggest established relationships, while low correlations may indicate under-researched areas ripe for further investigation.
- This methodology can guide future research by identifying promising disease-biomarker and disease-microorganism associations.
Related Concept Videos
Correlation and Causation
Correlation versus Causation
If the dependent variable increases or decreases when the independent variable increases, there is a positive or negative...
Correlations
Relationship Formation
Ending Relationships
Correlation
Two variables, for example, a and b, are said to be positively correlated if both variables move in the same direction. In other words, a positive correlation exists between two variables, a and b, if:
Theory of Attribution I: Correspondent Inference Theory

