Research And Implementation On Entity Alignment And Attribute Alignment

In recent years,with the development of the Internet technology,information on the Internet is increasing at an alarming speed.People can get information quickly and free ly from the Internet,and face the problem how to quickly and accurately obtain the information which they need.K nowledge group understands the query intentions of users by constructing knowledge,and finds the search results which users satify.Entity alignment and attribute alignmen is regarded as an important issue in the fields of knowledge group construction,web mining and intelligent information processing,and their technique can be applied into information retrieval,question answering system and automatic summarization and so on.This paper fouses on methods of entity alignment and attribute alignment based on the online encyclopedia.The task of entity alignment is to align entities which are from different websites and denote the same meanings.The task of attribute alignment is to merge the same attributes and the similar attributes which have the same meanings.For entity alignment,this paper proposes an entity alignment approach based on a multi-view fusion.The basic idea of this method is to align entities by two views including the view of free text and the view of infobox.The advantage of this method is to solve the entity alignment problem from multi-view,and consider the commonality and complementary principle of different views.For attribute alignment,this paper proposes a method based on Word2 vec.Its idea is to mine semantic web information by Word2 vec and the technology of distributed representation.Moreover,it integrates the similar attributes of entity.Its advantage is that this method effectively uses deep semantic information and short text knowledge,so as to enhance the effect of the attribute alignment.In this paper,the experimental datasets are from three online C hinese encyclopedia including Baidu,Hudong and Wikipedia,and include four different topics: tourist attractions,the protection of animals,people star and the countries in the world.The evaluation parameters of this paper are precision,recall and F-measure.The experiment results show that the entity alignment algorithm based on multi-view fusion outperforms the entity alignment algorithm based on a single view,and the performance of entity alignment algorithm based on multi-view fusion of BIRCH hierarchical clustering is better than that of the entity alignment algorithm based on multi-view fusion of LDA topic model and K-means clustering.Also,the result of the attribute alignment based on distributed representation is better than that of the attribute alignment based on similar distance.Therefore,the experimental results show that the approaches which are used for entity alignment and attribute alignment method is effective.Furthermore,the results of entity alignment and attribute alignment can be used to construct knowledge group,knowledge base and knowledge computing engine.