A Linear Dynamical System Model for Text

Low dimensional representations of words allow accurate NLP models to be trained on limited annotated data. While most representations ignore words’ local context, a natural way to induce context-depe