Journal international d'exploration de données biomédicales
Libre accès
ISSN: 2090-4924
ISSN: 2090-4924
Melinda Zhu
We describe a method for constructing a searchable database for medical doctors using information extracted from unstructured natural language texts on public websites. Specifically, we focus on biographical attributes such as the doctor’s medical school, undergraduate college and degree, age, medical specialties, publications on certain types of conditions (and their citation frequencies), associated media reports, etc. Ranking information for medical schools based on average MCAT scores and GPAs can be used as search parameters to provide for ranking of the search results. Citations for research publications and how often the doctor’s name is associated with particular types of medical conditions can also be used for ranking purposes. Absent of any personal knowledge of a particular doctor’s treatment outcomes, a patient looking for quality care can best be aided by ranked list of potential providers based on their educational backgrounds, experiences and knowledge of their specialties. Since we must collect the majority of our information from the Internet, which consists mostly of unstructured HTML based texts, finding specific information and categorizing them in a database requires natural language-based pattern recognition algorithms that can be learnt and associated with certain medical terms, as well as extract information on educational backgrounds and professional experiences. We argue that similar ideas can be applied to many other search tasks that can benefit from categorized databases built from the universe of unstructured web pages. We propose that a new type of web search engine can be designed using natural language processing techniques to mine extracted categorized information from unstructured texts to allow users to perform a variety of sophisticated searches that presently cannot be performed with current internet search engines.