Abstract: Aiming at the service demand of entity industry for scientific and technological resources,a text categorization model for distributed resource spatial based on multi-neural network fusion was put forward,which the service effect was taken as the criterion of resource text classification.A multivariate neural network path including word embedding layer,convolution layer,Bi-directional Gated Recurrent Unit layer(BIGRU),Attention layer and softmax layer were designed.On this basis,the strategy of demand-effect-resource classification was adopted to complete the mapping transformation from qualitative scientific and technological resource demand to quantitative resource service effect solution to qualitative resource output.Thus,the sensitive issues had been solved including the various forms for local and global semantic features of distributed scientific and technological resources,long-distance dependence of text,and difficult to accurately identify important resource information.Furthermore,the effect knowledge was acquired from distributed science and technology resource space quickly and accurately,the R&D efficiency and innovation ability of product development in real industry were improved.The feasibility and validity of the proposed method were verified by the patent data set of Wanfang Science and Technology Resources,which provided a new idea and means for more comprehensive mining of resource text features and on-demand service to the entity industry.

Key words: resource text classification, distributed resource space, multivariate neural network fusion, demand-effect-resource classification strategy, on-demand service

摘要： 针对实体产业对科技资源的服务需求,以服务效应作为资源文本分类标准,提出一种基于多元神经网络融合的分布式资源空间文本分类模型。设计了包含词嵌入层、卷积层、双向门控循环单元层、注意力机制层和softmax层的多元神经网络通路;在此基础上采用基于需求—效应—资源分类策略,完成了从定性科技资源需求到定量资源服务效应求解,再到定性科技资源输出的映射变换,重点解决了分布式科技资源局部和全局语义特征形式多样、文本长距离依赖特征显著、重要资源信息难以准确识别的问题,进而从分布式科技资源空间中快速准确地获取效应知识,提升实体产业产品研发效率和创新能力;通过万方专利科技资源数据集验证了所提方法的可行性和有效性,为更加全面地挖掘资源文本特征和按需服务实体产业提供了一种新的思路和手段。

关键词: 资源文本分类, 分布式资源空间, 多元神经网络融合, 需求&mdash, 效应&mdash, 资源分类策略, 按需服务