Using coherence-based measures to predict query difficulty

摘要: We investigate the potential of coherence-based scores to predict query difficulty. The coherence a document set associated with each word is used capture quality topic aspect. A simple score, QC-1, proposed that requires average contribution individual terms be high. Two further scores, QC-2 and QC-3, are developed by constraining QC- 1 in order semantic similarity among aspects. All three show correlation precision necessary make them good predictors Simple efficient, measures require no training data competitive language model-based clarity scores.