SOM-based Feature Map for Searching Similar Japanese Materials used in Junior/Senior High Schools

Abstract

Comparing two or more documents in Japanese lectures at junior/senior high schools is known to be effective for reading instructions to students. But it is difficult for teachers to continue these instructions, because there has not been any criteria to select suitable documents and it takes much time due to the massive materials. This study applies word2vec, doc2vec and Self-Organizing Map(SOM) so as to select the similar documents easily by visualizing the similarity on SOM-based 2-dimensional grid.