Schools without teachers? Workplaces without managers? Computers without users?
Imagine the chaos that would ensue if we sent children to school, with no teachers to supervise them. Or the wasted working hours if there were no managers to monitor and direct their employees. In the strange world of computer programmes, however, sometimes it is better to have no-one telling you what to do.
Prof. Pádraig Cunningham and Dr. Derek Greene from the Clique Research Cluster at UCD are studying unsupervised learning algorithms. These algorithms, which are complex combinations of computer instructions, can be used to explore and organise large amounts of information on a daily basis. News articles, blog posts, and Twitter feeds are common examples of the types of data that can be analysed by the algorithms. Because of the sheer amount of information involved, humans cannot possibly hope to manually analyse and organise this kind of data, or even regularly provide input to the software to aid in making decisions. So the software has to be effective in working independently from the user - through unsupervised machine learning techniques.
By learning without user guidance, these algorithms attempt to predict future trends for users on various topics of interest. For example, for a particular product, such as the Apple iPhone, the algorithms can analyse what kinds of comments and analysis appear online, and provide this information to the company for future use. The image below uses a machine learning system developed by Prof. Cunningham & Dr. Greene to automatically monitor the economic news reported by a number of popular online Irish news sources. The system has automatically found the most important words associated with the economic climate in Ireland.
Many organisations are interested in mining online opinions and trends to predict what will be the next iPod, the next Facebook, or the next Twitter. In particular, Polecat based in Dublin and the UK, is working with the UCD researchers to provide this kind of business intelligence to many of its clients. As part of this collaboration, Polecat has agreed a software licensing deal with UCD for a product developed by their researchers. The market they are competing in is predicted to be worth €2.8 billion in 2014.
