Research And Application Of Wordnet-Based Semantic Similarity Measurement

Calculation of semantic similarity is an important research content of natural language processing, and many measures have been proposed for the past few decades. These measures have been widely used in word sense disambiguation, detecting recognition errors in automatic speech transcripts, information extraction, spoken dialogue summarization, person name resolution, measuring the semantic similarity of texts, text classification, text clustering and other research fields. With the rapid development of internet, information processing becomes more and more important, especially the processing for text information, which becomes more complicated. Therefore, improving the accuracy of semantic similarity’s calculation is very important to the text information processing. With the further study and spread of the ontology, structured ontology has been proposed and applied to the calculation of semantic similarity, especially the semantic information of WordNet ontology, which has been widely used in measuring the semantic similarity.This thesis utilized depth and hyponyms of the concept nodes to calculate information content based on WordNet, and then proposed two new hybrid measures to calculate semantic similarity for the consideration of shortest path distance and IC semantic distance simultaneously. The experimental results show that the proposed methods are better than existing methods for the consideration of depth, distance and hyponyms simultaneously, and the results are more close to human judgment.This thesis applied the semantic similarity measurement to the matching of the semantic Web services. Because the proposed measurement can gain higher Pearson correlation coefficient, this measurement can judge the similarity of two words more accurately. Compared with the classic semantic Web services matching measurement, the proposed measurement can distinguish the inputs and outputs of the semantic Web services more effectively, so it can match the Web services accurately.Finally, this thesis designed and developed a semantic similarity calculation system, which was programed by Java with a GUI. The main function of this system is calculating the semantic similarity of two words with different measurements. All measurements described in this thesis can be applied in this system, and users can search word’s hypernyms or hyponyms with this system.