Event Causality Identification Using Conditional Random Field in Geriatric Care Domain

Event extraction is a key step in many text-mining applications such as question-answering, information extraction and summarization systems. In this study we used conditional random field (CRF) to extract causal events from PubMed articles related to Geriatric care. Abstracts of geriatric care domain were manually reviewed and categorized into 42 different sub domains. There are a total of 19, 677 sentences in the collected abstracts from PubMed, out of which 2, 856 sentences were selected and manually annotated with cause and effect events. The data set was then divided into training (2, 520), validation (252) and test (84) sentence sets. Features such as tokens, token categories, affixes, part of speech and shallow parser were used as inputs to the CRF model. A window of features before and after each token was used to determine its causal event label using CRF. A window of four features had the best performance with 84.6% precision, 87% recall, 85% and F-measure.