Analysis of Influence of Topic Models and Different External Corpus to Text Classification
Ying Na Zhu · Journal of Information and Computational Science · 2013
With the topic analysis models increasingly used in text categorization, many methods are developed to utilize topic analysis models to deal with the noises and sparseness of the text. It is very necessary to estimate the influence of the topic models to classification. And although exploiting external knowledge to enrich semantic of the text has achieved satisfactory results, choosing an appropriate universal corpus is still a knotty problem. In this study, we use topics extracted from texts by the LDA algorithm in two ways (using topics only and combining the topics and texts) to analyze the effect of topic models. And we also make use of different external corpus to value the importance of the external knowledge. The experimental results show that the topic models and the combination of different external datasets benefit categorization a lot.