Activity: Talk or presentation › Science to science
Extracting sentiments from text documents to forecast and compare it with economic indices is an attractive idea. We use a simple algorithm to analyze New York Times articles via sentiment analysis and compare our results to the Chicago FED National Activity Index (CFNAI) over a time horizon of 20 years. In order to derive sentiment scores from a particular ar- ticle we tag terms according to their conotated sentiment. Subsequently, we forecast the CFNAI movements via support vector machines (SVM) trained on the observed sentiment scores. We checked the forecasting ability of this approach by using the whole NYT articles, only the Business News, as well as using the changes of the sentiment scores to forecast the CFNAI.