PICO Classification Using Domain-Specific Features

This Paper proposes technique for extracting important information from clinical trial data. In order to extract important information PICO framework is used. PICO framework identifies the sentences in any given medical literature that fall under any one of the four categories: Participants/Problem (P), Intervention (I), Comparison (C), and Outcome (O). In this paper, machine learning classifiers are implemented to automatically detect PICO elements. The classification task is performed on the annotated clinical trial dataset. Instead of using a predefined word embedding like tf-idf, bag-of-word, or any other embedding domain-specific features are used as input to the machine learning classifiers. Domain-specific features were obtained using multiple entity extractors. These entity extractors extract entities related to a particular domain i.e., the biomedical domain. Extracted entities are converted into binary vectors and treated as features for the model. Our Proposed models perform better when compared to various strong baseline models.