Technical Articles | Project combat: How to construct Knowledge Atlas __https

This article is from Aliyun-yun-Habitat community, the original click here.

In practice, how to build a simple knowledge map, two versions, one starting from the 0 (start from scratch), one supplemented on cn-dbpedia basis, the MYSQL,POSTGRESQL,NEO4J database are all tried. I stumbled on the possibility of stepping on the pit do not know, welcome to discuss.

1. Cn-dbpedia Construction Process

The knowledge base can be divided into two types, one is the curated KBs represented by Freebase,yago2, which extracts a large number of entity and entity relationships from the knowledge base of Wikipedia and WordNet, such as a structured Wikipedia. The other is the Learning KBs, represented by Stanford Openie, and our school never-ending Language NELL (extracted), which extracts entity relationship triples directly from hundreds of billions of unstructured web pages.

The resulting knowledge is more diverse than freebase, but at the same time less accurate than curated KBs, since entity relationships and entities are more of a natural language form, such as "Obama was born in Honolulu." "Can be expressed as (" Obama "," is also born in "," Honolulu ").

Take Cn-dbpedia as an example to see how the knowledge map is generally constructed.

The above diagram is the Cn-dbpedia construction process and the system architecture respectively. The construction of the Knowledge Atlas is a vast project, from the big aspect, divides into the knowledge acquisition, the knowledge fusion, the knowledge verification, the knowledge computation and the application several parts, is the above structure diagram from the bottom to go up a flow, simple to walk this process.

Open the full text