基于深度学习的问答平台查询推荐研究  

Improving Online Q&A Service with Deep Learning

作　　者：丁恒 李映萱 Ding Heng;Li Yingxuan(School of Information Management,Central China Normal University,Wuhan 430079,China)

机构地区：[1]华中师范大学信息管理学院,武汉430079

出　　处：《数据分析与知识发现》2020年第10期37-46,共10页Data Analysis and Knowledge Discovery

基　　金：国家自然科学基金青年科学基金项目“基于深度语义表示和多文档摘要的学术文献自动综述研究”(项目编号:71904058);中央高校基本科研业务费资助项目“基于动态引文网络的人工智能算法演化路径研究”(项目编号:KJ02072020-0200)的研究成果之一。

摘　　要：【目的】针对社会化问答平台场景,构建深度神经网络模型,改善查询推荐的效果。【方法】以Yahoo Answers和Yahoo! L6为基础构建实验数据集,基于语义匹配矩阵、变长卷积层和多层感知机构建CNMNN神经网络模型,并与MQ2QC、IBLM、DRMM和MatchPyramid等基线进行了对比。【结果】对比MQ2QC、IBLM、DRMM、MatchPyramid这4种现有方法的最优效果,CNMNN模型在nDCG@5、nDCG@10、nDCG@20、MRR和MAP等相关性评价指标上的提升率分别为45.0%、38.7%、33.4%、34.8%和52.9%,在α-nDCG@5、α-nDCG@10、α-nDCG@20、ERR-IA@5、ERR-IA@10和ERR-IA@20等多样性指标上的提升率分别为31.5%、23.6%、25.5%、38.1%、36.9%和30.7%。【局限】尽管分析了多样性指标α-nDCG@k和ERR-IA@k,但是没有针对推荐结果提出进一步的多样化方法。【结论】CNMNN模型不仅可以计算查询和自然语言问句在短语级别的语义相关性,还避免了层次卷积操作导致的特征信号压缩问题。[Objective] This paper develops a neural network model to improve the online questioning and answering services. [Methods] First, we retrieved and constructed our experimental dataset from Yahoo Answers and Yahoo! L6 platform. Then, we proposed a neural network model(CNMNN) based on semantic matching matrix, variable-size convolutional layer, and multiple layer perceptron. Finally, we compared the results our model with the MQ2QC、 IBLM、 DRMM and MatchPyramid methods. [Results] The proposed model was45.0%, 38.7%, 33.4%, 34.8% and 52.9% higher than the best results on relevance metrics of nDCG@5, nDCG@10, nDCG@20, MRR and MAP. It also gained 31.5%, 23.6%, 25.5%, 38.1%, 36.9% and 30.7% improvements on diversity metrics of α-nDCG@5, α-nDCG@10, α-nDCG@20 and ERR-IA@5, ERR-IA@10 and ERR-IA@20.[Limitations] We did not include new method to further diversify the results. [Conclusions] The new CNMNN model can effectively calculate the semantic relevance between queries and natural language questions at phrase level. It also avoids the issue of feature signal compression due to hierarchical convolution operation.

关 键 词：查询推荐 深度学习 社会化问答 

分 类 号：TP393[自动化与计算机技术—计算机应用技术] G350[自动化与计算机技术—计算机科学与技术]