Integrating Medical Ontology and Pseudo Relevance Feedback For Medical Document Retrieval

The purpose of this thesis is to undertake and improve the accuracy of locating the relevant documents from a large amount of Electronic Medical Data (EMD). The unique goal of this research is to propose a new idea for using medical ontology to find an easy and more reliable approach for patients to have a better understanding of their diseases and also help doctors to find and further improve the possible methods of diagnosis and treatments. The empirical studies were based on the dataset provided by CLEF focused on health care data. In this research, I have used Information Retrieval to find and obtain relevant information within the large amount of data sets provided by CLEF. I then used ranking functionality on the Terrier platform to calculate and evaluate the matching documents in the collection of data sets. BM25 was used as the base normalization method to retrieve the results and Pseudo Relevance Feedback weighting model to retrieve the information regarding patients health history and medical records in order to find more accurate results. I then used Unified Medical Language System to develop indexing of the queries while searching on the Internet and looking for health related documents. UMLS software was actually used to link the computer system with the health and biomedical terms and vocabularies into classify tools; it works as a dictionary for the patients by translating the medical terms. Later I would like to work on using medical ontology to create a relationship between the documents regarding the medical data and my retrieved results.