Open Access Dissertation
Date of Award
Electrical and Computer Engineering
The online literature is an important source that helps people find the information. The quick increase of online literature makes the manual search process for the most relevant information a very time-consuming task and leads to sifting through many results to find the relevant ones. The existing search engines and online databases return a list of results that satisfy the user's search criteria. The list is often too long for the user to go through every hit if he/she does not exactly know what he/she wants or/and does not have time to review them one by one. My focus is on how to find biomedical literature in a fastest way. In this dissertation, I developed a biomedical literature search system that uses relevance feedback mechanism, fuzzy logic, text mining techniques and Unified Medical Language System. The system extracts and decodes information from the online biomedical documents and uses the extracted information to first filter unwanted documents and then ranks the related ones based on the user preferences. I used text mining techniques to extract PDF document features and used these features to filter unwanted documents with the help of fuzzy logic. The system extracts meaning and semantic relations between texts and calculates the similarity between documents using these relations. Moreover, I developed a fuzzy literature ranking method that uses fuzzy logic, text mining techniques and Unified Medical Language System. The ranking process is utilized based on fuzzy logic and Unified Medical Language System knowledge resources. The fuzzy ranking method uses semantic type and meaning concepts to map the relations between texts in documents. The relevance feedback-based biomedical literature search system is evaluated using a real biomedical data that created using dobutamine (drug name). The data set contains 1,099 original documents. To obtain coherent and reliable evaluation results, two physicians are involved in the system evaluation. Using (30-day mortality) as specific query, the retrieved result precision improves by 87.7% in three rounds, which shows the effectiveness of using relevance feedback, fuzzy logic and UMLS in the search process. Moreover, the fuzzy-based ranking method is evaluated in term of ranking the biomedical search result. Experiments show that the fuzzy-based ranking method improves the average ranking order accuracy by 3.35% and 29.55% as compared with UMLS meaning and semantic type methods respectively.
Alatrash, Massuod Hassan, "A Relevance Feedback-Based System For Quickly Narrowing Biomedical Literature Search Result" (2013). Wayne State University Dissertations. 827.