Gemini API Sentiment Analysis at Scale

Sentiment analysis with Gemini goes beyond positive/negative labels. You can extract nuanced sentiment, identify specific aspects being praised or criticised, detect emotion intensity, and classify feedback into actionable categories. ShopMax India analyses 10,000+ product reviews monthly to identify quality issues and surface top-rated products for promotion.

The key is a well-structured prompt that asks for specific output fields. Using JSON output mode ensures consistent machine-readable results that can be piped directly into your analytics pipeline without parsing errors.

The below example shows how ShopMax India classifies product reviews and extracts structured sentiment insights.

It gives the following output,

{
  "sentiment": "mixed",
  "score": 8,
  "aspects": [
    {"aspect": "battery", "sentiment": "positive", "comment": "lasts 2 full days"},
    {"aspect": "camera", "sentiment": "mixed", "comment": "decent but struggles in low light"},
    {"aspect": "build quality", "sentiment": "positive", "comment": "feels premium"}
  ],
  "highlights": "Battery life is amazing, lasts 2 full days",
  "complaints": "Camera struggles in low light",
  "recommend": true
}

For bulk analysis, process reviews in batches of 50 using the Batch API to reduce cost. Aggregate the results to find the most common complaints per product category and feed them into a weekly quality report for ShopMax India buyers.

It gives the following output,

Average satisfaction: 7.3/10
Top complaints: [('camera', 2), ('packaging', 1)]

Action: Flag camera quality to Samsung supplier contact.
Action: Review packaging process at Bangalore fulfilment centre.

Schedule this analysis as a Cloud Scheduler job every Monday morning. Write results to BigQuery and connect a Looker Studio dashboard for the ShopMax India merchandise team to track product sentiment trends week over week.