ReviewPulse

ReviewPulse, published by Monka31, reads a business’s review spreadsheet and groups praise and complaints in one dashboard. BERT and mDeBERTa handle sentiment and themes, while Qwen2.5 writes a diagnosis and reply drafts from the results. Filters by time, rating and theme help locate unanswered complaints. Replies remain drafts for human review, and the project warns that irony and mixed opinions can confuse the models.

Who it's for:Business owners and operations teams reviewing customer feedback, and people studying review-analysis workflows.

Before you start: The documented quick start uses Google Colab, an account and model-running resources; free GPU availability is not guaranteed. Input follows the supported review-export format. Confirm permission to use the data; the author warns that third-party collection may conflict with the source platform’s terms. We do not recommend posting unreviewed replies or publishing reports containing personal details. Code uses MIT, and documentation is mainly English.

Based on the project's public documentation. Not tested by AI Inspiration World.