MMR: A Multi-view Merge Representation model for Chemical-Disease relation extraction

Area of Science:

• Biomedical Informatics
• Natural Language Processing
• Computational Biology

Background:

• Chemical-Disease Relation (CDR) extraction is crucial for identifying semantic relationships between chemicals and diseases in biomedical texts.
• Effective representation of biomedical entities and their pairs is essential for accurate CDR extraction, posing challenges compared to general domain relation extraction.
• Existing methods may not fully capture the complex representations required for specialized biomedical texts.

Purpose of the Study:

• To propose a novel Multi-view Merge Representation (MMR) model for enhanced chemical-disease relation extraction.
• To improve the representation of both individual entities and entity pairs within biomedical documents.
• To achieve state-of-the-art performance on the CDR dataset.

Main Methods:

• Utilized prior knowledge and a pre-trained transformer encoder for capturing entity semantic representations.
• Employed U-Net and Graph Convolution Network layers to capture global entity-pair representations.
• Developed a merged representation for each entity pair to facilitate classification.

Main Results:

• The proposed Multi-view Merge Representation (MMR) model effectively captures entity and entity-pair representations.
• The model achieved state-of-the-art results on the CDR dataset.
• Demonstrated improved performance in chemical-disease relation extraction tasks.

Conclusions:

• The MMR model offers a robust approach for representing biomedical entities and their relationships.
• This advancement supports downstream applications in clinical diagnosis and drug discovery.
• The findings highlight the potential of multi-view representation learning in biomedical NLP.