A multi-approach to community question answering
作者：
Highlights：
• Simple first approach based on unsupervised Latent Semantic Indexing lead to state of the art results.
• Deep learning approaches help avoid the time consuming step of feature engineering.
• Pairwise approaches outperfom pointwise approaches.
• Using overlapping terms as inputs to neural networks improve both results and performance.
• Using pre-trained word embeddings reduce the overfitting of deep learning models.
摘要
•Simple first approach based on unsupervised Latent Semantic Indexing lead to state of the art results.•Deep learning approaches help avoid the time consuming step of feature engineering.•Pairwise approaches outperfom pointwise approaches.•Using overlapping terms as inputs to neural networks improve both results and performance.•Using pre-trained word embeddings reduce the overfitting of deep learning models.
论文关键词：Community question answering,Information retrieval,Arabic natural language processing,Learning to rank,Deep learning
论文评审过程：Received 12 May 2018, Revised 11 July 2019, Accepted 11 July 2019, Available online 12 July 2019, Version of Record 16 July 2019.
论文官网地址：https://doi.org/10.1016/j.eswa.2019.07.024
