學位論文
Permanent URI for this collectionhttp://rportal.lib.ntnu.edu.tw/handle/20.500.12235/73912
Browse
2 results
Search Results
Item Item Event Extraction for Gene Regulation Network Using Statistical and Semantic Approaches(2014) 班法; Bamfa CeesayGenic regulation networks are the primary study object in systems biology. They allow better understanding of the relationship between molecular mechanisms and cellular behavior. However, one of the bottlenecks in systems biology is the acquisition of an accurate genetic regulation network. In the recent years, the BioNLP community has produced systems for extracting genic interactions and Protein-Protein Interaction (PPI) from the literature. The sporulation network of the bacteria model for bacillus subtilis is very well studied. The automatic design of the gene regulation network is one of the main challenges in biology, because it is a crucial step forward in understanding the cellular regulation system. In this study, we present a description of a system on Gene Regulation Network (GRN) in bacteria and we use the data from the BioNLP’13 shared task (BIONLP-ST) on Event Extraction. For this work, we first propose a procedure to do biological event extraction combining a dependency graph-based method and a method using semantic analysis in Natural Language Processing (NLP). Then a second design, a statistical approach using Hidden Markov Model (HMM), is experimented. Dependency parsing is a significant and commonly used approach to finding out the dependency relationship between tokens in, for example, a sentence. We use dependency features to identify and classify our event trigger tokens using multi–class Support Vector Machine (SVMLight multiclass). However, the dependency features are not sufficient to give the semantic relationship between tokens with a sentence. Therefore, we develop a semantic analysis approach based on NLP techniques to capture more detail information and improve our result on event extraction. In our second design approach, we use a general statistical method via Markov’s logic instead of developing certain inferences and learning algorithms. Markov’s Model has achieved significant recognition in Natural Language Processing especially in the field of speech recognition. Our result shows that the graph-based approach obtains a better result on event extraction and produces a much better regulation network than the semantic analysis method. The combination of the two approaches has yet a much slightly better result than that with the individual approach. Moreover, the proposed statistical approach achieves a much better result than the combined and individual results of our graph-based and semantic analysis approaches.