Two-tier Architecture for Domain Specific Document Summarization Using Probabilistic Latent Semantic Analysis

In this research work we have proposed two-tier architecture for document summarization. This architecture minimizes the redundancy and boosts the information relevancy in the summary by applying Probabilistic Latent Semantic Analysis (PLSA) at two levels. It also enhances the summarizer’s speed by using Incremental Expectation Maximization algorithm for PLSA learning rather than Expectation Maximization. It starts with collecting number of topical information from multiple news portals and applies PLSA for single document Summarization. At next level PLSA is applied again in order to produce final summary but this time for multiple-document summarization. Here the two-tier stands for single and multiple-document summarization. In this paper we have performed a summarization experiment and we have also given a brief report of our experimental results. We experimented with variety of documents for several times and observed that results are generated in a matter of few seconds and moreover, the quality of the solution is fairly improved. To validate our results we have used the ROUGE metrics. The validation results devise the effectiveness of the proposed architecture.