Research On Relation Prediction Between Medical Ontology Terms By Fusing Machine Learning And Parallel Concept Sets

After entering the 21st century,medical domain ontologies have been widely used in medical information system construction,medical data processing and mining,etc.Their importance has been generally recognized by experts and scholars,and more and more scholars have started to participate in medical domain ontology construction and quality control and other related work.However,the traditional ontology construction work needs to be done manually by domain experts,which will not only take a long time but also consume great labor and material resources.Therefore,the establishment of an automated ontology building mechanism is urgently needed.Semantic relations are an important part of ontologies,and the prediction of semantic relations among ontology terms is an essential step in the study of automated ontology construction as a basic work of automated ontology construction.At present,the research on predicting semantic relationships between terms in medical ontologies at home and abroad is still limited to predicting relationships between single term pairs,and in most cases,at least one term in that term pair already exists in the medical ontology,and the research on predicting semantic relationships between two new terms and semantic relationships between terms in a new term set is very rare.In this paper,we firstly investigate the prediction method of semantic relationship between two new terms using machine learning models,and based on this,we further investigate the prediction method of semantic relationship between each term in the new term set.The specific work of this study is as follows.(1)In the research work on predicting the relationship between single term pairs,this paper first takes the term pairs as input features,learns the representation of the term pairs using textual representation methods and generates the corresponding feature matrices,and then fits these features using machine learning models to predict the relationship between single term pairs.(2)In the research work on automatic prediction of relationships among terms in a term set,this study first pairs the terms in the term set two by two,and all the term pairs generated by the pairing form a set of candidate term pairs,and then uses a machine learning model to predict the relationships among the term pairs in the set of candidate term pairs.The experimental results showed that this method produces a large number of false positive results due to the high number of candidate term pairs.To solve this problem,this study proposes a PCS-based algorithm that then incorporates a machine learning model to predict the relationship between each candidate term pair.The experimental results show that the effective screening of candidate term pairs by the PCS-based algorithm effectively improves the accuracy of the prediction results of the relationships among terms in the term set.The experimental results of fusing machine learning models with PCSbased algorithmic strategies on the task of automatic prediction of interterm relationships in term sets show that the algorithm proposed in this paper performs well on ontologies,with an average F1 value of 0.79 for100 case test results,indicating that the algorithm can predict the relationships existing among terms in a term set to a large extent,and thus construct the semantic structure of the term set.This can improve the efficiency of ontology construction by domain experts and provides a new idea for the study of automatic ontology construction.