Stephan Mandt

Dynamic word embeddings

2017 · International Conference on Machine Learning

We present a probabilistic language model for time-stamped text data which tracks the semantic evolution of individual words over time. The model represents words and contexts by latent trajectories in an embedding space. At each …