LDA Based Related Word Detection in Advertising

Abstract

In this paper, we propose a new method for related word detection in Advertising by combining LDA topic model and word co-occurrence. We use a corpus of BaiduBaike, which is a Chinese Encyclopedia, to calculate the word co-occurrence. Words allocation on topics driven by LDA is used to sort the related words glossary which is obtained by the traditional co-occurrence procedure. We evaluate our method on advertising related word recognition, and the experiments result shows that the method is feasible.