Design And Implementation Of Scholar Portrait System Based On Network Representation Learning

With the wide application of big data technology in academia,academic big data has gradually become an important data support for scientific research,and its analysis has also become an important topic.The center of academic research is scholars.Using academic big data to build portraits for scholars and introducing various machine learning algorithms to provide various services for scientific research and innovation has become a current hot research direction.This project relies on the actual subject of my internship unit,and the subject is mainly based on academic scientific and technological data to build an open innovation platform.This paper mainly studies the construction of scholar portrait,the construction of scholar portrait system and the specific application of scholar portrait in the platform based on the construction of academic knowledge base,combined with user portrait technology.First of all,in order to get through the data isolation and data quality problems between the existing academic platforms,the system obtains academic data from various sources as much as possible,and integrates the data from the third-party partner institution database,the public academic data set and the academic homepage of the university.Academic data is then cleaned and normalized to improve the accuracy of subsequent analyses.Based on this,this paper designs a scholar data collection module and a metadata management module to facilitate data supervision and build a complete academic knowledge base.Secondly,referring to the existing construction ideas of user portraits,this paper generates shallow dimension scholars portraits from the perspective of demographics,and uses text mining and network representation learning methods to generate deep dimension scholars portraits.The shallow-level dimensional portraits include: basic information cards of scholars,statistics of academic achievements,portfolios of academic achievements,and various academic relationship maps.The deep dimension profile includes: scholarly research hot words and scholarly representation vectors.In order to obtain a more representative scholar vector,this paper introduces a network representation learning algorithm to construct a scholar’s academic text information and co-authorship network.This method takes the scholars as the node,through the characteristics of node text and maximize the adjacent structure of node nodes and the intellectual information of the network,to obtain higher-order nodes to characterize as a scholar’s characteristic vector.The system defines the dimensions of the portrait through the label management module,statistics and calculations of the data,and set the label version record function and historical retrospective function,so that the label can better adapt to the academic knowledge base at different stages.Scholars’ portrait custom module supports different portraits of individual scholars to facilitate timely adjusting portrait data and portrait construction methods.Third,this article introduces the use of scholars ’portraits in the platform,including scholars’ search classification,interested scholars recommendation,and scholar academic homepage generation.Among them,the scholars’ intervention module can manually adjust the scholars’ search results.By defining the weights of some scholars in the results of the search,modify their ranking scores to avoid the deviation of the retrieval results caused by problems such as data errors.Finally,the above modules are combined to build a scholar portrait system.The author mainly completed the system requirements analysis,database design,detailed design and front-end and back-end development work.At this stage,it has served the academic science and technology platform and has achieved good results.In addition,this system solves the problem of data isolation,obtains more representative scholars’ characteristics vector,and completes the various applications of scholars’ portraits on the actual platform,making the scholars’ portrait construction process more processed.