Using Semantic ‎PSO ‎Clustering Approach for Automatic Text Summarization

Sentence Clustering is often used as a first step in document summarization to find redundant information. Number of clusters, type and quality of them can have important roles in the automatic text summarization. Also the similarity criteria have effective roles in coverage of principal and significant sections of a text. In this research, it is tried to use a new approach in the text summarization problem based on PSO (Particle Swarm Optimization). Similarity of sentences is calculated based on semantic correlations and it is contrary to similarity based on word co-occurrences. In this research, number of clusters is not considered as a predefined parameter and it is tried to find the optimal cluster numbers. The proposed system is evaluated on a large dataset of sports news. The results show that the output of this system is more efficient and accurate than similar approaches.

Keywords:

Summarization , Clustering , PSO algorithm , Number of clusters , Similarity , Vectors containing terms , K-Means