Research Of Word Similarity Computing In Ontology Automatic Generation

With the rapid developing of network technology,more and more information resources of every domain in society can be acquired from network by people..It is becoming a hot issue that how to retrieve valuable information on web quickly and accurately home and abroad.The similarity of vocabulary terms is an important question in the information retrieval.In the machine translation,automatic response system,meaning disambiguation,natural language processing and other fileds term similarity calculation has a wide range applications.As the emergence of semantic web,word similarity computing also plays an important role in the Ontology integration and semantic information retrieval.Based on the above background,firstly this article introduces the theory and concept of ontology,and then analyzes the present research situation of word similarity and several of representative arithmetic on word similarity.The article presents an improvement method based-on word vector of similarity calculation.The dictionary is constituted by a series of words as well as their related interpretation. Some designated vocabulary act as the key word.This key word is explained by its explanatory note,and the words in the explanatory note are also explained by their own explanatory notes.Consequently,a vocabulary network hierarchy is formed. Then calculating probability of each word which appears in the corresponding explanation note.These probability data is saved to matrix document(frequency matrix).And calculates word vector through the formula(I-aA)C=(1-a)A.The similarity of words is defined as the probability that a key word is estimated from explanatory note of another key word.This similarity expresses how closely a key word represents the explanatory note of another key word.The greater their value,the closer the two words,and vice versa.These words come from the word vector file, and the value of the similarity is stored as vector files finally.Then this paper verified this method using sample data,and the experimental results show that the method of similarity calculation is feasible.Such similarity calculation method laid the foundation for words classification.