Latent topic estimation based on events in a document

Recently, several latent topic model-based methods such as LSI, pLSI, and LDA have been widely used for text analysis. However, those methods basically assign topics to words, and therefore the relationship between words in a document is not considered. Considering this, we propose a latent topic extraction method which assigns topics to events that represent the relation between words in a document. There are several ways to express events, and the accuracy of estimating latent topics differs depending on the definition of an event. Therefore, we propose several event types and examine which event type works well to estimate latent topics in a document with a common document retrieval task. Furthermore, as an application of our proposed method, we also show a multi-document summarization based on latent topics.