It is important to understand "worked better". In practice, I would trade off using simpler methods for few points increase in precision/recall/your choice of metric.
You can try my site Last10K.com which uses sentiment analysis to find & highlight positive & negative remarks in lengthy 10K/Q filings. Here's an example from Burlington Stores' 10K this week where we found 60 positive and 15 negative remarks by their management team: