NMTF(Normalizing Mapping Training Framework) operates by mapping initial cluster centers and then iteratively training points to clusters base on the proximate cluster center and updating cluster centers. We regard finding good cluster centers as a normalizing parameter estimation problem then constructing the parameters of other normalizing models yields a space of novel clustering methods. In this paper we propose the idea using abstract of a text to members of a data partition in place of estimation of cluster centers. The method can accurately reconstruct meaning representation group used to generate a given data set.
展开▼
