| Traditional Chinese Medicine(TCM)is one of the quintessence of China,which has developed for thousands of years.As an important carrier of TCM,clinical records reflect the applying of li-fa-fang-yao principle,contain rich experience of famous doctors and have the function of "declaring the past,learn for future".However,due to diversity of style,mixture of ancient and modern text and poor standardization,it is hard to label,organize and mine knowledges from clinical records.Thus,there is no professional knowledge service system for clinical record now.Under the background of China Knowledge Center for Engineering Science and Technology,this thesis aims at analyzing,mining and displaying inherent knowledge from clinical records,and focuses on research of the key technology applications and both design and implementation of the clinical record system.The main works are as'follows:1)For lacking of clinical record labels,we present a novel topic model for TCM clinical record classification,which extracts concept words and separates different types of concepts.It could extract more coherent topics and better topic distribution,which increases the accuracy of clinical record classification.2)To ensure the uniqueness of each clinical record,rule based Simhash algorithm was applied to remove duplicate clinical records and ensure the richness and diversity of clinical record set.3)When analyzing and finding closeted classic recipes,recipes are extracted from clinical records by CNN at the size of sentence,reflecting the rules of drug usage.4)In order to improve the precision of service data,we present a crowdsourcing scheme.Incorrect data is corrected by user submitting opinions and experts reviewing.Most voting and Dawid&Skene algorithms are used for quality control.5)Based on these research,clinical record system is designed and implemented,providing functions of searching,classify browsing,reading,recipe analysis,clinical record analysis and error correction on line. |
PDF Full Text Request