Towards an Obesity-Cancer Knowledge Base: Biomedical Entity Identification and Relation Detection

Area of Science:

• Biomedical Informatics
• Computational Biology
• Public Health

Background:

• Obesity is a significant risk factor for numerous cancers and chronic diseases.
• Accessible, reliable health information empowers patients and improves health outcomes.
• Current online information on obesity and cancer is fragmented and lacks formal structure.

Purpose of the Study:

• To develop methods for constructing a formal knowledge base on obesity and cancer.
• To improve the organization and delivery of quality health information for consumers.
• To address limitations in existing ontologies for automatic knowledge base construction.

Main Methods:

• Implemented Named-Entity Recognition (NER) using linguistic and statistical approaches to extract biomedical entities from scholarly articles.
• Developed a relation detection method based on statistical features from sentences to identify links between biomedical entities.
• Leveraged state-of-the-art techniques to surpass existing results in NER.

Main Results:

• The proposed NER methods achieved state-of-the-art performance.
• The relation detection method demonstrated high accuracy (99.3%) and F-measure (0.993).
• The developed techniques are foundational for building a comprehensive obesity-cancer knowledge base.

Conclusions:

• The developed NER and relation detection methods are effective for building a structured obesity-cancer knowledge base.
• This work facilitates better organization and accessibility of vital health information.
• Future work will focus on expanding the knowledge base to aid public health initiatives.