Description: I am experiencing issues with the sentiment analysis provided by the Google Gemini API. Specifically, the API’s safety filters seem to be affecting the sentiment analysis results, leading to neutral sentiments being assigned to content that should be classified differently.
Context:
- Use Case: Our application requires accurate sentiment analysis to categorize video content based on its title and description.
- Issue: The API’s safety filters are impacting the accuracy of sentiment analysis. For instance, a video description that should be classified as “negative” was classified as “neutral.”
Example:
- Prompt: “What sentiment value does this video have based on its title ‘ghost death incident’ and description ‘famous ghost death incident case horrifying’?”
- Response: The result was ‘neutral’ due to safety filters.