조요한 연구실

Accurate mortality prediction is an important task in intensive care units in order to channel prompt care to patients in the most critical condition and to reduce nurses’ alarm fatigue. Nursing notes carry valuable information in this regard, but nothing has been reported about the effectiveness of temporal analysis of nursing notes in mortality prediction tasks. We propose a time series model that uncovers the temporal dynamics of patients’ underlying states from nursing notes. The effectiveness of this information in mortality prediction is examined for mortality prediction for five different time spans ranging from one day to one year. Our experiments show that the model captures both patient states and their temporal dynamics that have a strong correlation with patient mortality. The results also show that incorporating temporal information improves performance in long-term mortality prediction, but has no significant effect in short-term prediction. User-generated reviews on the Web contain sentiments about detailed aspects of products and services. However, most of the reviews are plain text and thus require much effort to obtain information about relevant details. In this paper, we tackle the problem of automatically discovering what aspects are evaluated in reviews and how sentiments for different aspects are expressed. We first propose Sentence-LDA (SLDA), a probabilistic generative model that assumes all words in a single sentence are generated from one aspect. We then extend SLDA to Aspect and Sentiment Unification Model (ASUM), which incorporates aspect and sentiment together to model sentiments toward different aspects. ASUM discovers pairs of {aspect, sentiment} which we call senti-aspects. We applied SLDA and ASUM to reviews of electronic devices and res