Title: Fine-grained, aspect-based semantic sentiment analysis on news for economic forecasting and nowcasting
Authors: Sergio Consoli - National Research Council of Italy (CNR) (Italy) [presenting]
Sebastiano Manzan - European Commission (Italy)
Luca Barbaglia - European Commission Joint Research Centre (Italy)
Abstract: News are a promising nowcasting and forecasting tool since they describe current economic events and the expectations of economic agents about the future. In particular understanding the sentiment embedded in current economic news may provide additional signals to improve forecasts of economic models and produce more accurate predictions. Recent work in economics on the application of sentiment analysis from social media and news generally suffers from: (i) a limited scope of historical financial news available; (ii) analysis of short texts only (e.g. usually tweets or news headlines); and (iii) use of basic text analysis techniques. We provide an overview on the development of a fine-grained, aspect-based sentiment analysis approach. The method is based on advanced natural language processing and is able to recognize the exact entity to which the sentiment aspect is expressed within entire, long text articles. The approach is unsupervised since it relies on external lexical resources to associate a polarity score to each concept. We describe our novel method and report some preliminary findings on historical economic news which consider a long time period of 25 years. Our analysis provides evidence on the usefulness of considering news sentiment as an additional instrument in the forecasting toolset of macroeconomists.