Overview
This analysis examines Google Play Store reviews for SafeBoda — a mobility app operating in Africa. The investigation uncovers customer sentiments, common complaints, and language patterns using R text mining.
Data Source
Dataset scraped from Google Play Store using Beautiful Soup (Python), then loaded into R for analysis. The original data contained 14,290 reviews with columns including review content, ratings, and timestamps.
Key Findings
Most Common Words
After filtering stop words and app-specific terms, the analysis identified top words revealing what customers discuss most frequently.
Sentiment Analysis
Positive sentiment leaders: "perfect" (4.92 avg rating), "fantastic" (4.91), "excellent" (4.90)
Negative drivers: "error" associated with lowest avg rating (1.83), followed by "version" (2.03) and "phone" (2.18)
Temporal Trends
Between 2017 and 2020, positive sentiment and customer trust increased significantly while negative sentiment declined — indicating improving product satisfaction over time.
Bigram Analysis
Common word pairs reveal specific issues: "takes forever," "unknown error," and "code verification" emerge as frequent phrases that pinpoint concrete pain points.
Negation Patterns
Phrases like "not good" and "not happy" often mask negative sentiment when analysed without context — a key challenge in naive sentiment models.
Methodology
Analysis employed tidytext, NRC lexicon, and AFINN sentiment dictionaries to categorise emotional language and track sentiment evolution over time.