PROJECT TITLE :

Biomedical Relation Extraction With Knowledge Graph-Based Recommendations

ABSTRACT:

Biomedical Relation Extraction (RE) systems search for and categorize relations between biomedical entities in order to improve our understanding of biological and medical procedures. Deep Learning techniques are used in the majority of today's most advanced computer systems, primarily to focus on relations between entities that are of the same type, such as proteins or pharmacological substances. On the other hand, these systems are mostly limited to what they can directly identify on the text, and they ignore specialized domain knowledge bases such as ontologies that formalize and integrate biomedical information and are typically structured as direct acyclic graphs. On the other hand, the importance of integrating Knowledge Graphs (KGs) in order to add additional features to items has already been demonstrated by Recommendation Systems that are based on Knowledge Graphs. The typical definition of a system's users is individuals and the things they consume, which can range from movies to books and include anything else people watch or read and then categorize based on how satisfied they were. This work proposes the incorporation of KGs into biomedical RE by means of a recommendation model in order to further improve their range of action. We came up with the idea for a new RE system and gave it the name K-BiOnt. This system was created by combining a state-of-the-art baseline deep biomedical RE system with a state-of-the-art existing KG-based recommendation system. According to the findings of our research, including recommendations derived from KG-based recommendation boosts the system's capacity to recognize genuine connections between entities that the baseline deep RE model was unable to derive from the text.