VoiceScope VOC Customer Feedback Insight

Problem addressed

VOC feedback often exists as scattered tickets, reviews, or spreadsheets. Reading raw rows makes it hard to identify recurring pain points, positive signals, or a stable tagging basis. This skill targets existing CSV, .xlsx, .xlsm, or line-based TXT feedback files and turns raw text into structured, auditable, exportable analysis results.

How it works

It acts as an entry point and orchestration layer, delegating authentication, task execution, LLM analysis, and persistence to the VoiceScope backend. Core capabilities include:
- Viewpoint extraction + clustering: extract row-level viewpoint snippets with sentiment, then cluster them into pos, neg, and neu Top-K clusters; export the official insight Excel file.
- Batch tagging: classify each feedback item after confirming a taxonomy_id, reporting hits and misses.
- Taxonomy generation and validation: generate a draft taxonomy from samples, with structural checks and optional --llm semantic review.
- General analysis columns: analyze a selected column row by row for sentiment, keyword, summary, scenario, or action.

The key step is to check login status, preview the target column, and get explicit confirmation before upload. This prevents running row-billed analysis on the wrong column.

Boundaries

It is suitable for summarizing, clustering, tagging, and row-level analysis of textual feedback, but it does not replace platform-side visualizations such as word clouds, taxonomy insights, or sentiment dashboards. Legacy .xls files are not supported, and the internal signature_keywords fingerprint in clustering is not an external keyword result.