Distribution and Analysis of Negative Headlines in Nigerian News: A Scalable Computational Framework

This study presents a scalable computational framework to analyze negative sentiment and linguistic patterns in 1,238 news headlines from a Punch, a Nigeria newspaper. The news headlines were randomly selected from a dataset originally crawled in 2018, and this analysis focused on 786 negative headlines across 40 topics. The essence of this work is not to present a current view of the Nigeria landscape but rather to explore a functional framework for analysing negative news usable regardless of content date. Using NLP techniques including topic modeling and sentiment analysis and other ad-hoc processing, we developed a scalable framework utilising parallel processing for efficiency. Statistical analysis of output from the applied experiment reveals a skewed distribution, with Topic 13 (fraud) as an outlier (6.1% of coverage) and a Gini coefficient of 0.48. Linguistic analysis identifies “death” (54 occurrences) and “murder” (negativity score 0.375) as key negative concepts, with Topics 6, 7, 13, and 35 reflecting violence and political turmoil. Visualizations elucidate patterns, offering insights for media studies and policy. The findings reveal a disproportionate media focus on violence and fraud, which can skew public perception locally and internationally and suggests policy makers should engage more proactively with lingering issues.