SenticLens (Sentic Labcast by Simerjit Kaur Nath)

SenticLens is a production oriented sentiment analytics framework that processes reviews from Reddit, the Apple App Store, and Google Play, analyzing polarity, intensity, and emotion via SenticNet while auditing symbolic inconsistencies with SenticCheck. Using 4,000 reviews across McDonald’s, Netflix, Nike, and CeraVe, we evaluate three open weight LLMs (Gemma 2 9B, Llama 3.1 8B, and Mistral Nemo 12B) for polarity emotion coherence under the Hourglass model. Across 24,000 audited records, incoherence rises from 18.49% in synthetic reviews to 38.16% in real reviews, demonstrating robust ambiguity amplification across models and brands. SenticLens thus provides both deployment ready review analytics and a benchmark for affective reliability. https://sentic.net/senticlens.pdf