Named Entity Recognition from Financial Press Releases.

This paper explores a previous model's use of named entity recognition to predict the changes in stock prices from financial news. Detecting company mentions in the articles is crucial for this task, and we modified these methods to gain additional mentions. We first expanded upon the rules of the named entity recognition from the original model. We also incorporated coreference resolution and modified an existing toolkit to be compatible with our specific domain. After these two adjustments, the number of instances captured increased significantly. Although this did not necessarily improve the overall prediction performance, the results give us an opportunity to explore reasons why the scores stayed around the same, and a full analysis will allow us to achieve our goals.