基于关系图卷积网络的答案相关评论检索算法

A Review Retrieval Algorithm Based on the Relational Graph Convolutional Network

针对基于检索方法的产品相关的问答系统在编码问题、参考答案和评论语句时,单一地提取文本的句子级语义信息而忽略了不同类型的文本之间的多种语义关系,提出了一个基于关系图卷积网络的答案相关评论检索模型。该模型通过在不同类型的文本之间显式地建立多种语义关系来增强相关信息的表达。相关的实验结果显示,该模型优于所有的对比模型,在实际的评论检索情景中也能够准确地检索出与问题相关的评论。

For the product related question answering based on retrieval methods,when encoding questions,reference an-swers,and reviews sentences,it only extracts the sentence-level semantic information of the text and ignores the multiple semantic relationships between different types of text.A review retrieval model based on the relational graph convolutional network is pro-posed to enhance the representation of sentences by explicitly establishing semantic associations among different types of sentences.The experimental results show that this model outperforms all comparison models,and can accurately retrieve the reviews related to the given question.