1: \begin{abstract}

2: Functional Distributional Semantics

3: is a framework that aims to learn, from text,

4: semantic representations which can be interpreted in terms of truth.

5: Here we make two contributions to this framework.

6: The first is to show how a type of logical inference can be performed

7: by evaluating conditional probabilities.

8: The second is to make these calculations tractable

9: by means of a variational approximation.

10: This approximation also enables faster convergence during training,

11: allowing us to close the gap with state-of-the-art vector space models

12: when evaluating on semantic similarity.

13: We demonstrate promising performance on two tasks.

14: \end{abstract}

15: