Autoencoding Pixies: Amortised Variational Inference with Graph Convolutions for Functional Distributional Semantics
- Submitting institution
-
University of Cambridge
- Unit of assessment
- 11 - Computer Science and Informatics
- Output identifier
- 12825
- Type
- E - Conference contribution
- DOI
-
10.18653/v1/2020.acl-main.367
- Title of conference / published proceedings
- Proceedings of the 58th Annual Meeting of the Association for Computational Linguistics
- First page
- 3982
- Volume
- 0
- Issue
- -
- ISSN
- 0000-0000
- Open access status
- Compliant
- Month of publication
- July
- Year of publication
- 2020
- URL
-
-
- Supplementary information
-
-
- Request cross-referral to
- -
- Output has been delayed by COVID-19
- No
- COVID-19 affected output statement
- -
- Forensic science
- No
- Criminology
- No
- Interdisciplinary
- No
- Number of additional authors
-
0
- Research group(s)
-
-
- Citation count
- 0
- Proposed double-weighted
- No
- Reserve for an output with double weighting
- No
- Additional information
- This introduces the use of a graph-convolutional neural network to perform approximate inference of latent variables efficiently. This is a significant development for the framework of Functional Distributional Semantics. The new inference algorithm has been used to advance the state of the art on two tasks in computational semantics.
- Author contribution statement
- -
- Non-English
- No
- English abstract
- -