Question utility: a novel static ranking of question search

In this paper, we propose a notion of 'question utility' for studying usefulness of questions and show how question utility can be integrated into question search as static ranking. To measure question utility, we examine three methods: (a) a method of employing the language model to estimate the probability that a question is generated from a question collection and then using the probability as question utility; (b) a method of using the LexRank algorithm to evaluate centrality of questions and then using the centrality as question utility; and (c) the combination of (a) and (b). To use question utility in question search, we employ a log linear model for combining relevance score in question search and utility score regarding question utility. Our experimental results with the questions about 'travel' from Yahoo! Answers show that question utility can be effective in boosting up ranks of generally useful questions.