Abstract:
Significant advances are being made in the field of biomedicine. The biomedical researchers present their research finally in the form of text. It has become a challenge for biomedical researchers to search for the latest discoveries of protein functions and interactions and required information in the ever increasing published literature. Text mining thus becomes helpful in conversion of this information into structured database format. In this report, we propose the method of function phrase chunking to extract protein functions and interaction phrase chunking for protein-protein interactions from biomedical literature. The proposed methods can be used to assist database curators in interpreting protein function and interaction terms and to aid bio-medical researchers in searching protein functions and protein-protein interactions from biomedical text literature. The experimental results on more than 0.12 million abstracts of protein functions and 0.17 million abstracts of protein-protein interactions shows the effectiveness of our system. The high precision and recall ranges of 87.6%-70% and 70.6%-59% respectively of our protein function extraction system and similarly the high precision aInd recall ranges of 94%-85% and 90%-73% respectively of protein-protein interaction extraction system indicate the outperforming efficiency of our system in comparison with existing systems
