An empirical assessment of best-answer prediction models in technical Q&A sites