AI Sentiment Automation Engine — Sandbox

Train a lightweight sentiment model from labelled examples, upload unlabelled feedback, preview every dataset before modelling, auto-label every row, and ask analysis questions about the results. This version is designed to feel fast inside the portfolio while mirroring the Flask workflow.

Best workflow: load training data → choose text and label columns → train → load unlabelled data → choose text column → predict → ask the assistant for insights.

Column guide: the text column contains the feedback or comments to analyse. The label column contains known labels such as positive, neutral, or negative.

Upload labelled CSVUse your own training data with at least one text column and one sentiment label column.

Text column for training

Pick the column containing the actual sentences, reviews, complaints, survey responses, or beneficiary comments.

Label column for training

Pick the column containing the known sentiment class. Accepted labels are positive, neutral, and negative.

Prediction data: this dataset does not need a sentiment label. The engine keeps your original columns and appends sentiment, confidence, and class probabilities.

Upload unlabelled CSVReviews, support tickets, call notes, donor feedback, survey comments, complaints, or social media responses.

Text column to analyse

Pick the column containing the text you want the trained model to score. The output dashboard updates after prediction.

Your question

Optional Groq API key

Groq model

Security note: in this static portfolio sandbox, the key is used only in your current browser session and is not stored by this page. For production, route Groq calls through the included Flask backend using an environment variable.

Smart routing: the local assistant handles summaries, drivers, confidence, and ROI. If your question asks for strategy, root cause, executive narrative, comparison, or prioritisation, the sandbox will recommend using Groq for deeper analysis.

No model trained yet. Load sample training data and click Train Domain Model.

Rows loaded

Columns loaded

—

Current preview source

Not trained

Model status

Status
No dataset loaded yet.

Rows labelled

—

Training accuracy

—

Average confidence

Negative items

Positive

Neutral

Negative

Labelled Output

Status
Train the model and run predictions to generate labelled output.

Welcome.

Load data first. I will then explain sentiment distribution, negative themes, confidence, business impact, and recommended actions from the processed output.

AI Sentiment Automation Engine — Sandbox

Output Dashboard

Model Results

Labelled Output

Analysis Assistant

AI Sentiment Automation Engine — Sandbox

Output Dashboard

Dataset Preview

Model Results

Labelled Output

Analysis Assistant