| With the high-speed evolution of information technology,the occurrence of Knowledge Graph has aroused extensive attention,which is used in information retrieve,personalized recommendation and QA system.Nowadays,the content of most General Knowledge Graph comes from the infobox of online encyclopedias with high quality,which is under scale and restricts the development of information technology.Considering HTML tables in the Internet are rich in knowledge and semi-structured HTML tables have higher quality than unstructured text,this paper will focus on knowledge base augmentation research based on HTML tables and construct the system.However,it is a challenging problem to obtain HTML tables with knowledge,correctly and effectively extracting triples to augment knowledge graph.Above all,Internet resources are unreliable,which needs a way to get high-quality HTML tables.Secondly,HTML tables have different types,which needs corresponding extraction methods.Thirdly,the contain of HTML tables shows differently from knowledge graph,so that we need to match the HTML tables to knowledge graph.This paper mainly studies a way of knowledge base augmentation based on HTML tables and implements the system.The main work of this article is as follows:(1)It researches and implements a knowledge base augmentation system based on HTML tables,which is able to augment CN-DBpedia.The results show that the system is able to extract and fuse millions of triples for CN-DBpedia.(2)It designs a data collection framework for HTML tables.For the diversity of tables,this paper uses the table features and sample set to train a table recognition model to get HTML tables that contain knowledge.For the low credibility of Internet resources,this paper utilizes the web features and sample set to train a credible web recognition model to obtain the web resources.(3)It proposes a knowledge extraction and fusion framework over HTML tables for knowledge base augmentation,which provides one-stop service on dealing with various HTML tables,extracting knowledge and integrating into knowledge base. |
