Microblogs and social media platforms are now considered among the most popular forms of online communication. Through a platform like Twitter, much information reflecting people's opinions and attitudes is published and shared among users on a daily basis. This has recently brought great opportunities to companies interested in tracking and monitoring the reputation of their brands and businesses, and to policy makers and politicians to support their assessment of public opinions about their policies or political issues.
A wide range of approaches to sentiment analysis on Twitter, and other similar microblogging platforms, have been recently built. Most of these approaches rely mainly on the presence of affect words or syntactic structures that explicitly and unambiguously reflect sentiment (e.g., “great”, “terrible”). However, these approaches are semantically weak, that is, they do not account for the semantics of words when detecting their sentiment in text. This is problematic since the sentiment of words, in many cases, is associated with their semantics, either along the context they occur within (e.g., “great” is negative in the context “pain”) or the conceptual meaning associated with the words (e.g., “Ebola“ is negative when its associated semantic concept is “Virus“).
This thesis investigates the role of words' semantics in sentiment analysis of microblogs, aiming mainly at addressing the above problem. In particular, Twitter is used as a case study of microblogging platforms to investigate whether capturing the sentiment of words with respect to their semantics leads to more accurate sentiment analysis models on Twitter. To this end, several approaches are proposed in this thesis for extracting and incorporating two types of word semantics for sentiment analysis: contextual semantics (i.e., semantics captured from words' co-occurrences) and conceptual semantics (i.e., semantics extracted from external knowledge sources).
Experiments are conducted with both types of semantics by assessing their impact in three popular sentiment analysis tasks on Twitter; entity-level sentiment analysis, tweet-level sentiment analysis and context-sensitive sentiment lexicon adaptation. Evaluation under each sentiment analysis task includes several sentiment lexicons, and up to 9 Twitter datasets of different characteristics, as well as comparing against several state-of-the-art sentiment analysis approaches widely used in the literature.
The findings from this body of work demonstrate the value of using semantics in sentiment analysis on Twitter. The proposed approaches, which consider words' semantics for sentiment analysis at both, entity and tweet levels, surpass non-semantic approaches in most datasets.