Presented herein is a novel model for similar question ranking withincollaborative question answer platforms. The presented approach integrates aregression stage to relate topics derived from questions to those derived fromquestion-answer pairs. This helps to avoid problems caused by the differencesin vocabulary used within questions and answers, and the tendency for questionsto be shorter than answers. The performance of the model is shown to outperformtranslation methods and topic modelling (without regression) on severalreal-world datasets.
展开▼
