Research On Literature Mining Algorithms And System Integration For Biochemical Network

With the rapid development of biology, biological information grows at exponential rate. As one of the most important way to display academic achievements, biomedical literature is large in number and growth speed is far greater than any other discipline. The overwhelming amount makes it impossible for researchers to keep up with the state of the art result manually. Therefore, a automatic literature mining system become a necessary tool for researchers. In this article, we do some key algorithm researches on literature mining area with respect to biochemical network. The main content includes the following aspects:1. Although there are many available literature mining tools, many of them targeting only some specific function. We put forwards an general solution for biomedical literature mining—BNLitMiner. BNLitMiner is implemented by interrogating some important algorithms and developing an integrated system (BioIMIS). BioIMIS is based on J2EE architecture and possesses the virtues of robustness, extensibility, portability2. Traditional Bayesian methods in the biomedical literature are often in favor of longer documents. For this problem, we gave an improved Bayesian algorithm (LRABIB) by taking document length factor into consideration. The results show that LRABIB has some improvement on recall and precision compared with the general Bayesian method and reduces the complexity of the algorithm.3. Support Vector Machine (SVM) classification in the biomedical literature often lacks the combine with domain knowledge. For this problem, we integrated the gene ontology information into the SVM classification model and proposed GO-SVM algorithm. The experimental results show that GO-SVM algorithm has a higher F-Score compared with the general SVM classification.To obtain the knowledge of biochemical networks from biomedical literature through literature mining methods, not only has great value in theoretical perspective, but also has broad prospects. In this paper, we conducted some research on several key literature mining algorithms and the application of system integration. How to import more domain knowledge will be the future direction of our work.