摘要
语义匹配对许多自然语言处理任务至关重要,诸如信息检索中信息匹配、问答系统中问题和答案的匹配等.基于语义的匹配,即通过提取文本内在语义进行匹配度计算,是目前自然语言处理领域研究的热点.本文提出一种基于深度神经网络的文本语义匹配模型--多粒度语义交叉模型,从语义匹配的角度来解决文本匹配问题.模型首先通过循环神经网络获取短文本不同粒度的语义表示,然后从两个短文本不同粒度的语义交互信息中提取它们语义匹配信息,从而计算两个短文本的语义匹配度.实验表明,本文提出的基于多粒度语义交叉模型在短文本匹配上表现出较好的计算效果.
Semantic matching is crucial for many natural language processing tasks,such as information matching in information retrieval,and the matching of questions and answers in question answering systems. Semantic-based matching,that is,calculating the matching degree by extracting the intrinsic semantics of the texts,is currently the focus of research. This paper proposes a text semantic matching model based on deep neural network a multi-granularity semantic cross model,to solve the text matching problem from the perspective of semantic matching. The model first obtains the semantic representations of different granularities of short texts through the Recurrent Neural Network,and then extracts the semantic matching information from the semantic interaction information of two short texts,so as to calculate the semantic matching degree of the two short texts. Experiments show that the multi-granularity semantic cross model shows good performance and computational results on short text matching.
作者
吴少洪
彭敦陆
苑威威
陈章
刘丛
WU Shao-hong;PENG Dun-lu;YUAN Wei-wei;CHEN Zhang;LIU Cong(School of Optical-Electrical and Computer Engineering,University of Shanghai for Science and Technology,Shanghai 200093,China)
出处
《小型微型计算机系统》
CSCD
北大核心
2019年第6期1148-1152,共5页
Journal of Chinese Computer Systems
基金
国家自然科学基金项目(61772342、61703278)资助
关键词
语义匹配
自然语言处理
多粒度
语义交叉
深度神经网络
semantic matching
natural language processing
multi-granularity
semantic cross
deep neural network
