Dr Nina Dethlefs
Deep Text Generation from a Knowledge Graph
People Involved
Project Description
The objective of this project to develop a text generator that can (a) generate coherent and grammatical text for a specific domain from a knowledge graph of relevant facts, and (b) transfer the learnt linguistic representations to new domains that have a similar function but can have different semantic fields. This project is designed as an MSc by research degree in Computer Science and is sponsored by Stanford-based start-up company Diffbot, who are providing us with access to their Knowledge Graph of facts as well as noisy training data. The main methodology is based on the MemN2N model, a sequence-to-sequence recurrent neural network that learns to condition an output sequence of words on an input sequence of semantic fields. An external memory component helps align portions of noisy text fragments with fields from the knowledge graph, so that no human engineering is required in model training.
| Status | Project Complete |
|---|---|
| Value | £21,119.00 |
| Project Dates | Oct 1, 2017 - Sep 30, 2019 |
You might also like
The EPSRC Centre for Doctoral Training on Offshore Wind Energy and the Environment Apr 1, 2019 - Sep 30, 2027
The move to cleaner economic growth is one of the greatest industrial opportunities of our time; a key industrial challenge for the 21st Century is to secure a resilient renewable energy supply. The offshore wind (OSW) industry underpins UK solutions...
Read More about The EPSRC Centre for Doctoral Training on Offshore Wind Energy and the Environment.
CAMS PostDoc Nov 1, 2019 - Jan 31, 2022
Pollution of freshwaters from industry, agriculture and urban settings is a major pressure on human and ecological health. Effective and robust monitoring of water quality is vital to safeguard supplies but traditional monitoring has relied on expens...
Read More about CAMS PostDoc.
KTP - Spencer Aug 3, 2020 - Aug 3, 2022
Use natural language processing and artificial intelligence to mine data, analyse and extract information to improve workflows and support informed and effective decision making.
Secure and Safe Multi-Robot Systems Jan 1, 2021 - Dec 31, 2023
European strategy and research roadmap documents emphasise the significant societal and economic benefits coming from robotic and autonomous systems. Multi-Robot Systems (MRS) comprise distributed and interconnected robotic teams that can carry out t...
Read More about Secure and Safe Multi-Robot Systems.
KTP - RB Oct 1, 2021 - Sep 30, 2023
This project developed novel algorithms to improve stability testing of new product formulations and accelerate their launch from initial conception and design through to testing and manufacturing., enhancing Reckitt’s capacity to react to customer f...
Read More about KTP - RB.