Text Segmentation Based on PLSA-TextTiling Model

Text segmentation is very important for many fields including information retrieval, summarization, language modeling, anaphora resolution and so on. Text segmentation based on PLSA-TextTiling associates different latent topic swith observable pairs of word and sentence. In the experiments, the whole sentences are taken as elementary blocks. PLSA model is used to calculated similarity metric basing on the idea of TestTiling and several approaches to discovering boundaries are tried. The results show the Pμ value is 0.87, which is better than that of other algorithms of text segmentation.