Model in Distress: Sentiment Analysis on French Synthetic Social Media
This work addresses the high cost of annotated data, scarcity of evaluation sets, and privacy concerns in multilingual social media analysis for customer feedback.
The authors developed a synthetic data generation pipeline for customer distress detection in French public transportation, producing 1.7 million synthetic tweets from a small seed corpus. Their 600M-parameter reasoners achieved 77-79% accuracy on human-annotated data, matching or exceeding SOTA models while reducing annotation costs and preserving privacy.
Automated analysis of customer feedback on social media is hindered by three challenges: the high cost of annotated training data, the scarcity of evaluation sets, especially in multilingual settings, and privacy concerns that prevent data sharing and reproducibility. We address these issues by developing a generalizable synthetic data generation pipeline applied to a case study on customer distress detection in French public transportation. Our approach utilizes backtranslation with fine-tuned models to generate 1.7 million synthetic tweets from a small seed corpus, complemented by synthetic reasoning traces. We train 600M-parameter reasoners with English and French reasoning that achieve 77-79% accuracy on human-annotated evaluation data, matching or exceeding SOTA proprietary LLMs and specialized encoders. Beyond reducing annotation costs, our pipeline preserves privacy by eliminating the exposure of sensitive user data. Our methodology can be adopted for other use cases and languages.