Publication | Conference on Neural Information Processing Systems 2019

Relational Graph Representation Learning for Open-Domain Question Answering

“Knowledge graphs” formalize knowledge about a domain of interest. They can contain multimodal information about entities and represent complex interactions among objects or concepts within a domain. Unlike ontologies, knowledge graphs are compatible with deep learning framework and can automatically predict missing knowledge. Knowledge graph technology is vastly used in software companies such as Google to augment their search results, Amazon to give better recommendations, and also in more specialized domains such as healthcare. Autodesk has also a knowledge graph called “Design Graph” which represents the interactions between parts of an assembly. This research can show how to use knowledge graphs such as Design Graph to augment downstream tasks such as classification, recommendation, or question-answering. Moreover, considering the domain of this project (natural language processing), it can help with applications that include textual data such as automatic extraction of Building Codes.

Download publication


Relational Graph Representation Learning for Open-Domain Question Answering

Salvatore Vivona, Kaveh Hassani

Conference on Neural Information Processing Systems 2019

We introduce a relational graph neural network with bi-directional attention mechanism and hierarchical representation learning for open-domain question answering task. Our model can learn contextual representation by jointly learning and updating the query, knowledge graph, and document representations. The experiments suggest that our model achieves state-of-the-art on the WebQuestionsSP benchmark.

Related Resources



BOP-Elites: A Bayesian Optimisation Approach to Quality Diversity Search with Black-Box descriptor functions

An algorithm that efficiently tackles expensive black-box optimization…



SLiMe: Segment Like Me

We explore leveraging extensive vision-language models for segmenting…



COIL: Constrained Optimization in Workshop on Learned Latent Space

Constrained optimization problems can be difficult because their…



A Learning Approach to Robot-Agnostic Force-Guided High Precision Assembly

In this work we propose a learning approach to high-precision robotic…

Get in touch

Something pique your interest? Get in touch if you’d like to learn more about Autodesk Research, our projects, people, and potential collaboration opportunities.

Contact us