Text mining or data mining is a knowledge discovery tool which is referred to the process of extracting interesting and non-trivial patterns from a database of unstructured texts. Here, we present a new machine learning system to mine biological data sets (text data/scientific literature) to understand relations between two genes (two terms) in a scientific text. The system mimics human intelligence and accurately determine the relations between two genes/proteins. We manually curated literature data sets using deep curation to generate training set. Furthermore, our prediction results were validated with the help of experts to generate confidence to use our system in different real time situations. Next the system was made automated so that people across the world can determine relations between two or more molecules in a text using support vector machines. This semi-automated system is frequently applied by our team to write reviews on a given topic. For example, our team was able to screen and mine over 36000 papers to write a review on molecular docking tools. In 2016, our team were able to reconstruct obesity molecular network using this system(Jaisri et al 2016, Plos One).