Sentiment analysis using a lexicon-based approach in Lisbon, Portugal.

Area of Science:

• Computational Social Science
• Spatial Analysis
• Natural Language Processing

Background:

• The proliferation of digital data from users in diverse emotional states presents opportunities and challenges for spatial research.
• Analyzing large volumes of user-generated data necessitates advanced techniques like sentiment analysis to extract meaningful insights.
• Understanding public sentiment in urban spaces is crucial for urban planning and social research.

Purpose of the Study:

• To identify urban locations in Lisbon exhibiting predominantly positive or negative public sentiment.
• To leverage sentiment analysis on social media data for spatial research applications.
• To explore the correlation between specific types of urban spaces and prevailing emotional tones.

Main Methods:

• Utilized the Canadian National Research Council (NRC) Sentiment and Emotion Lexicon (EmoLex) for text analysis.
• Analyzed data sourced from the social media platform Twitter (now X).
• Mapped sentiment scores to geographical locations within Lisbon.

Main Results:

• Portuguese users exhibit positive sentiment in areas associated with leisure and consumption, including museums, event venues, gardens, shopping centers, stores, and restaurants.
• The study identified a bias in negative sentiment scoring due to lexicon challenges in contextual understanding.
• Specific words, even in neutral contexts, were sometimes incorrectly assigned negative scores (e.g., 'war', 'terminal').

Conclusions:

• Leisure and consumption spaces in Lisbon are associated with positive public sentiment.
• The accuracy of sentiment analysis, particularly for negative emotions, can be influenced by lexicon limitations and contextual ambiguity.
• Future research should address contextual nuances in sentiment lexicons for more precise spatial sentiment analysis.