Predictive performance comparisons of different feature extraction methods in a financial column corpus

This work concerns the processing of a corpus made up of a financial weekly column. Specifically, we focused on document-level index extraction and textual feature extraction. Moreover, some feature extraction methods had been compared to evaluate their predictive capacity. Results confirm the hypothesis that vectors derived from word embedding do not improve the predictive power compared to other feature extraction methods but remain a fundamental resource for capturing semantics in texts.