The advent of social media sites offers many new sources of data to analyse and use in novel ways. We have now available a vast amount of information about people, brands and corporations – what they say and what is said about them, who they connect to and who connects with them, how they respond to events and become authorities on events. Wars, sport, elections, floods, famines can all be viewed through the lenses of social media activity. Below we describe some of the main aspects that we have been exploring the field. For more details on the ongoing research, please refer to the publications page.
We have been applying sentiment analysis techniques to analyse the influence and reach of brands and people. This analysis most likely involves the following steps: gathering data from a social media forum (e.g. Twitter), categorising and clustering the data based on features and sentiment; and identifying trends in the data.
Generating and visualising predictions in social media
With the richness of data available on social media, much focus has been placed on making useful predictions regarding data, places and people. We have been interested in investigating how best to generate a prediction based on the existing sources of evidence and secondly how best to explain the prediction visually. Analysis of user behaviour, when given these explanations, are also undertaken.
Personalised user interaction in social media systems
Retrieval and organisation of data and information from social media systems (e.g. Facebook and Twitter) is a very active area of research which poses several challenges to researchers and developers (scale of data, heterogeneity, real-time retrieval). We have been researching and developing suitable user interaction models based on users’ multiple needs. Here, our goal is to identify task at hand, users’ interests, users’ interaction style/model and to present information and paths to information based on those tasks and interaction styles.
Analysis and prediction of growth in dynamic graphs on social media
The increased use of social media has led to the growth of research into such graphs. One of the open questions is to predict the growth of such graphs and the spread of information through these graphs. In this work, graphs are built based on interconnections between people (and clusters of people). Properties relating to the growth of such graphs are captured. Data mining approaches are used to predict new connections in the graph and also to predict what type of behaviour leads to activity by other posters. To attempt to predict this, we mine the relationships between graph properties, user activities and the resulting activities by other users.