In recent years, the development of artificial intelligence is in full swing . The application of artificial intelligence forms the threshold of enterprises , Data forms the competitiveness of enterprises , Knowledge mapping is a key artificial intelligence technology , It's about making data knowledge , It can make machines understand the human world better .
In China's domestic knowledge mapping technology , Peking University is ahead of us . PKU knowledge map database gStore Many applications have been launched . 10 month 31 A source theory technology salon was held in Zhongguancun in Junxin technology and Beida data laboratory on Japan. The application scenarios of map recognition were analyzed and discussed , Invite to Peking University 、 tencent 、 The technical experts of ant group and Junxin technology share their experience of knowledge mapping technology .
Salon activity site 50 Multiple people and online 400 Multiplayer participation . There are participants from all walks of life ： There is insurance 、 Finance 、 military 、 Knowledge payment 、 education 、 Retail, etc. . Participants share through the guests of this salon , Combined with their own fields, this paper discusses the application scenarios of many knowledge maps .
JunXin Technology CTO Luo Fei shares the theme ： What is the knowledge graph . The concept of knowledge map has been popularized for the field participants . Artificial intelligence technology is divided into three stages ： Computational intelligence -> Perceptual intelligence -> Cognitive intelligence . Knowledge mapping is the key technology in cognitive intelligence stage . Junxin technology will gStore Graph technology is used in natural language processing , Give Way AI The dialogue is more precise .
Tencent senior researcher Li Youhuan shares the theme ： The practical experience and thinking of graph technology re industry scenario . Three new information sources and related technologies of knowledge map are shared , They are the association structure information 、 Based on Association propagation aggregation attribute information , And the nearest neighbor vector represents information . Besides , Li Youhuan also introduced the application scenarios of the three major information in the industry , Including the detection of fraud gangs , Financial risk control arbitrage mining and so on .
Ant group algorithm expert Yang Shuo shared ： The practice of large scale graph learning in ants . Ant financial intelligence team developed a spatiotemporal map , Hierarchical heterogeneous graph neural network and other technologies , The accuracy of risk identification model is further improved . And developed a graph learning engine AGL, Support large scale graph learning in a large number of ant business scenarios
Li Wenjie, postdoctoral of Peking University, shared ： Construction of medical knowledge map based on knowledge map , This paper introduces the development process and Prospect of medical knowledge map , The introduction is based on “ disease - disease - drug ” The construction process and methods of the medical knowledge map based on , And the typical application cases based on the knowledge map of medicine .
HelloCode System architect Li Muyi shared ：ThinkAPI The map of knowledge in and ThinkPHP6 New characteristics . Explain how to make good use of the service combination related to knowledge map in the application development layer ThInkAPI And ThinkPHP Develop products quickly .
There are also business representatives present ：HelloCode Children programming COO Guo Xin , Beijing Xinliu Huijie Technology Co., Ltd CEO Zhang Xinghui .
gStore Graph database , It was a lab project in Peking University , follow BSD Open source agreement , After years of development and project deployment , It's pushing for gStore Social application of , The salon is gStore The first step in social application . Enhance people's discussion of knowledge map through Technology Salon , Let more people apply the knowledge map to their own work , And let more people participate in gStore Open source project contribution .
gStore The advantages of graph database system ：
For the first time, the academic community proposed to construct the query engine of knowledge map storage based on subgraph matching ; gStore System core published 3 Top papers in database field (SIGMOD,VLDB,VLDB J), The highest citation rate of a single article exceeds 200 Time ;
8 Years of research and development , Supported by the national key R & D program , System 10 Ten thousand lines of code are written independently , Own from “ The paper - patent - Software copyright ” The overall intellectual property system , The system is completely domestic and controllable ;
Complete product level development and practical deployment , Has been applied to finance 、 Big government data 、 Intelligent question answering robot 、 Telecom fraud detection and public security big data projects ;
Performance is remarkable
China Software Testing Center test results show that , Distributed gStore Scalability of graph data supporting 10 billion edge scale , And the average query response time is seconds . Academic colleagues at home and abroad 、 The evaluation reports of several system application parties and authoritative testing organizations show that : The performance obviously surpasses the foreign similar graph database product , It is more prominent in large-scale graph data and complex query .
本文为[Source theory Technology Salon]所创，转载请带上原文链接，感谢