Reading news from the internet and predicting the stock market

speaker: Yue Zhang, SUTD event: DataScience.SG ** event driven predictions E=(O_1, P, O_2, T) O = object P = action T = timestamp open information extraction ** categorize each event as either positive or negative on the stock using SVM and neural network ** evaluation accuracy and MCC metric ** conclusion events > bag of words non linear model > linear model ** represent an event as a vector neural tensor network ** take into account long term effect of events convolution neural network and max pooling for long term and mid term events