Chinese Document Clustering Using Self-Organizing Map-Based on Botanical Document Warehouse

The exponential growth of information has made an overflow situation in the sea of information. It had created difficulties in the search for information. An efficient method to organize the query of information and assist users’ navigation is therefore particularly important. In this paper, we applied Self-Organizing Map (SOM) algorithm to cluster Chinese botanical documents onto a two-dimensional map. Each botanical document has been regarded as bags of words, and transferred into plain text respectively. We applied term frequency and inverse term frequency to extract key terms from documents as the input of SOM. 892 Chinese botanical documents have been projected onto a 2D map to assist users’ navigation. In our experimental results, the lowest recall was 0.71 for Polygonaceae documents and the highest recall rate was 0.94 for Amaranthaceae documents. The lowest precision rate was 0.81 for Umbelliferae documents, and the highest precision rate was one hundred percent for Convolvulaceae and Cruciferae documents.