engleski

Spatio-Temporal Word Embeddings

Latent Topic Models are a powerful tool for classifying, clustering and representing documents. For dynamic collections, we need to model both temporal and topical evolution. In this work, we present two models that use a continuous time distribution and that represent words as trajectories in a continuous space. We perform experiments on Twitter data and compare our model to some state of the art models in this field, but also a simpler baseline, all of this shows the potential of our model,

distributed representation ; topic detection ; time ; clustering ; micro-blogs ;

nije evidentirano