Towards Using Scientific Publications to Automatically Extract Information on Rare Diseases

A small percentage of the population is afflicted by what is called an orphan or a rare disease. All over the world, there are about several thousand of these diseases. When adding up together all the individuals who are affected, it amounts for up to 10% of the US population. Scientific works on these diseases are often poorly financed due to the lack of potential markets for a treatment, which means for patients and clinicians a very limited and scattered access to vital information. To contribute addressing this issue, we present in this paper a new software tool for automating the extraction of information related to rare diseases from scientific publications. More precisely, our contribution consists in a new method of extracting automatically symptoms of these diseases from research papers exploiting a Named Entity Recognition (NER) algorithm based on the numerical statistic Term Frequency - Inverse Document Frequency (TF-IDF). The proposed tool has been tested using PubMed Central (PMC) database.