The programs made available here are only a small sample of the work of the NLG group. In many cases they are on a very small scale, and may have been constructed simply to demonstrate some very limited facility. Usually our research projects produce much larger and more complex software, which it is not practical to place here for online experimentation.
Please do try out our demo software and let us know what you think about it.
The STANDUP project developed interactive software to enable children, particularly those with communication disabilities, to explore language by making simple jokes (punning riddles). The Joking Computer project developed this software into a public exhibit, including a version which runs on the web. The demo here is that web version. Go to demo
This is a simple data-to-text system. Starting from a table of numbers which represent predicted pollen levels in different regions of the country, it generates a paragraph of text summarising that information. On the demo page, you can see past summaries (both human-written and computer-generated) for specific dates, or you can provide some numerical pollen levels for the various regions and see how the computer program describes your data.
This demo allows you to enter information relevant to your cardiac health, such as your age and whether or not you smoke, and to see the percentage risk for various health outcomes. If you select the Summary tab, you will see a computer-generated text outlining the salient facts in the diagnosis. The other tabs provide numeric or visual representations (you may need to widen your browser window to see the full display). Using the left panel enter some medical information about yourself ... then look at the pages of information concerning your potential for heart disease!
The STOP project developed software which would, starting from certain facts about a person, generate a short letter which tried to persuade that specific person to stop smoking. The demo version (which is slightly simpler than the version used in full trials) allows you to enter information about your smoking habits (just invent some facts if you're not a smoker!) and see the letter that would be sent to you.
The Sum-Time project created a data-to-text system which started from numerical information representing a prediction about weather conditions, and generated a short paragraph summarising that forecast in English. The demo allows you to enter wind-speed predictions for a 24-hour period and then see how that would be described.