Research Explores New Methods for Opinion Mining Social Content

A research paper published recently in the Fifteenth International AAAI Conference on Web and Social Media used social opinion datasets as a basis to explore new methods for opinion mining online content.  The paper was written by ADAPT researcher, Keith Cortis along with Professor Brian Davis of the SFI ADAPT Centre and in it they used three high quality social opinion datasets related to Malta’s annual Government Budgets of 2018, 2019 and 2020 to explore the sentiment and opinions in the content.  

Opinion mining or sentiment analysis is a popular and valuable research area where user-generated content extracted from social sites and commenting sections on newswires can give insights about citizens’ perceptions and needs, and society’s problems at large.  According to Eurostat (2019 statistics), 13% of individuals living in Malta post opinions on civic or political issues via websites, such as blogs and social networks. 

The researchers worked with three datasets which contained over 6,000 online posts of user-generated content in English and/or Maltese, gathered from newswires and social networking services.  This provided a voice to the citizens who use social media platforms to make their opinions known and/or provide  feedback about any particular measure announced by the Government, whether it is tax related, industry specific, or any other social initiative. 

The datasets are a valuable resource for developing Opinion Mining tools and Language Technologies, and can be used as a baseline for assessing the state-of-the-art and for developing new advanced analytical methods for Opinion Mining.  

Opinion mining is considered a challenging Natural Language Processing (NLP) problem, especially when applied on social data. This evolving field, also known as Social Opinion Mining (SOM), is tasked with the identification of  several opinion dimensions.  In this case the content was annotated for multiple opinion dimensions, namely subjectivity, sentiment polarity, emotion, sarcasm and irony, and in terms of negation, topic and language. 

The three datasets provide a valuable resource for developing opinion mining tools that gather political and socio-economic insights from user-generated content in Malta’s two official languages, Maltese and English.  These can be used by the Government of Malta for policy formulation, policy-making, decision-making, and decision-taking.  Moreover, their use can support similar initiatives in other countries, studies in the socio-economic domain and other application areas, such as Politics, Finance, Marketing, Advertising, Sales and Education.

Full research paper is available here.