11.05.2009. The Validation and Business Applications Group attached to the Artificial Intelligence Department at the Universidad Politécnica de Madrid’s School of Computing has developed a new intelligent text mining system that restyles traditional opinion poll systems.
The system has been successfully applied in research developed jointly with the Universidad de Zaragoza based on a questionnaire conceived to find out university students’ opinion about Gran Escala, a project to build Europe’s biggest leisure complex in Aragon’s Monegros district.
The Zaragoza Multicriteria Decision-Making Group (GDMZ) from the Universidad de Zaragoza set up a website for this purpose. The website was only accessible to the administrator and the student census selected to give their opinion. This census was the basis for the experiment.
Through this website students gave a total of 332 opinions on the pros and cons of the Gran Escala project. Most students rated the project positively, and only a small minority considered it to be harmful to the district.
The really important thing about the survey, though, was not the actual results, but the research conducted on the results. This research was part of what is known as e-cognocracy, the democracy of the knowledge society.
Students gave two different types of opinions: some opinions were expressed as ratings, where they scored the opportunities and threats of the project, and others were entered in a forum as texts expressing their opinions on Gran Escala more freely and informally.
Taking the opinions rating Gran Escala, the GDMZ applied numerical and statistical techniques to find out the opinion of the student body on the project. This way, they were able to identify not only their preferences but also the arguments supporting their opinions.
While the GDMZ worked on the numerical and statistical research, the Validation and Business Applications Group (VAI) at the UPM’s School of Computing designed intelligent text mining techniques to analyse the messages that the students had entered in the forum. The results output by this system were very similar to the ones indicated by the numerical and statistical techniques.
As explained in a paper to be presented at the Second World Summit on the Knowledge Society to be held on 16, 17 and 18 September at Crete (Greece), the match between the results is 72.31% in the case of positive responses in favour of Gran Escala and 40% in the case of responses contrary to the project going ahead at Los Monegros.
According to VAI’s director, Jesús Cardeñosa, this result opens the door to a new way of interpreting citizen opinion, making it a possible complementary alternative to traditional survey-taking systems.
Traditional opinion poll systems are based on quantitative techniques that measure the preferences of a group of people and establish the reasons why people think what they do about something. Systems like these offer respondents a range of previously drafted responses from which citizens have to choose the one that best represents their opinions.
The new technique for researching opinions is different. This system allows citizens to express their opinion about a project or political option through freely drafted text, which is a significant innovation with respect to traditional methods. Responses are not conditioned and can also be further specified.
The intelligent text mining system developed by the VAI analyses these texts, defines grammars and extracts linguistic patterns, offering a more thorough and less biased view of citizens’ opinions. The system not only discovered the same preferences as expressed in the ratings, but also found that positive opinions are more precise than negative opinions.
A potential use for these techniques, which can be extended to more important public affairs, is voting intention forecasting. Here they can be used in conjunction with traditional systems, which are based primarily on the arguments underlying the opinions expressed through surveys with set responses.
This item on other websites
Gran Scala 11.05.2009