Yahoo! for Amazon: Sentiment Extraction from Small Talk on the Web
Management Science
Patterns of temporal variation in online media
Proceedings of the fourth ACM international conference on Web search and data mining
| Hi-index | 0.00 |
There is now a small but growing literature showing some relationship between sentiment contained within blogs, online news article and message boards and price movements in financial markets. Typically, researchers use keyword searching to find financially relevant messages, then rate them in terms of their how positive or negative the sentiment they contain is in relation to prices. Through an exploratory analysis of the statistical nature of word frequency movements on Twitter, we highlight some issues with this approach and define how a sentiment variable could be constructed to generate well specified linear regression models. We then address a second issue of how to model time. Current research has used units of a day or week for both sentiment and price series. There is no discussion in the literature in this area as to what the best unit of time might be, or indeed, if there is a weekly topology to sentiment price correlations. We present two models which explore how these factors affect sentiment-price correlations. Finally we present results correlating financial sentiment on Twitter to the price of the Standard and Poor's Index of 500 Leading Shares. We report both contemporaneous (R squared values up to 0.35) and predictive correlations (R squared values up to 0.27) between our sentiment metric and prices. Scale and weekly topology both appear significant factors that would benefit inclusion in future models.