Data mining social media opinionsNovember 29th, 2013 in Technology / Internet
A European collaboration has analysed thousands of microblogging updates to help them develop an opinion detector for data mining the social media lode and extracting nuggets of information that could be gold dust for policy makers, marketing departments and others looking for emerging trends and attitudes. Details are to be published in the International Journal of Electronic Business.
Many people are now inclined and able to share their opinions widely thanks to social media, on microblogging sites including Twitter and online social networks such as Facebook. Whether anybody takes any notice of those opinions is a moot point. We see endlessly strident and highly offensive comments on virtually every Youtube clip. News stories, particularly those on emotive subjects such as abortion, religion, evolution, climate change, twerking and selfies, are "trolled" narcissistically.
But, for those in the world of commerce and in particular the marketing wing of many organizations, all that commentary is not meaningless, it is a deep lode of information to be mined. Within the gems unearthed one might find the collective opinion on almost any product or service, the trends, the fancies of the early adopters and the likes and dislikes of the masses. Marketing mavens everywhere are looking for ways to dust off these gems and to polish them up for the consumption of their sales and advertising teams. One such methods of ways of panning the Twitterhood for precious nuggets of insight that could mean the difference between a marginal profit margin or a company marginalized on the whim of publicity has been developed by a research team in Greece.
Informatics experts Michail Salampasis of the Alexander Technological Educational Institute of Thessaloniki, in Sindos, Anastasia Giachanou of the University of Macedonia, in Thessaloniki working with Georgios Paltoglou of the School of Technology, at the University of Wolverhampton in the UK, have analysed hundreds of thousands of microblogging messages containing comments, sentiments and opinions about food and brand products.
Social networking on sites like Facebook and Google+ and microblogging services, such as Twitter, coupled with our 24/7 always connected via mobile or broadband attitude means that countless people cannot escape the opinions of others or of sharing their own ever wider. The team's system harvested millions of tweets and used a computer algorithm to automatically extract the sentiment from those tweets.
"Our results provide strong indications that given the use of such services by millions of users, they can play a key role in supporting and enhancing important business processes," the team says. They suggest that key aspects of the world of modern marketing are not so different from those that existed before online social media - company-to-customer relationship management, brand image building and Word-of-Mouth (WoM) branding - but today the rate at which information might be exchanged is so much faster than it ever was before. Moreover, a positive message that goes viral can lead to an enormous sales boom whereas a deleterious comment adopted as a true reflection of a given product by the many will lead to a bust that could lead to the subduction of a product or even a company very rapidly.
The team's analysis of well known brands as well as world affairs demonstrates how data mining twitter can spot shifting opinion on fast-food outlets, wars or potentially even famine and flood. "We believe that the amount of information contained in microblogging websites makes them an invaluable source of data for continuous monitoring ...using opinion mining and sentiment analysis techniques," the team concludes.
More information: "Using social media for continuous monitoring and mining of consumer behaviour" in Int. J. Electronic Business, 2013, 11, 85-96
Provided by Inderscience Publishers
"Data mining social media opinions." November 29th, 2013. http://phys.org/news/2013-11-social-media-opinions.html