Student Feedback Tone Analyzer

How Student Feedback Tone Analyzer works

Press Load model once: your browser downloads MobileBERT sentiment classifier (MediaPipe) (33 MB) from this site and keeps it in its cache. Then add your text and press Analyze locally. The model runs on this device, so nothing you enter is sent to a server. Until the model is loaded, or if your device cannot run it, the tool falls back to a simple keyword check and labels the result as the fallback.

MobileBERT sentiment classifier (MediaPipe) is a compact BERT model fine-tuned on the Stanford Sentiment Treebank (SST-2), a set of movie-review sentences labelled positive or negative. The tool scores your whole text, then every sentence on its own.

A sentence counts as clearly positive or negative when the model is at least 80% sure. If clearly positive and clearly negative sentences both appear, the overall result is Mixed; otherwise it follows the whole-text score. The most negative and most positive sentences are quoted so you can see what drove the result.

Worked example

The model's output for the built-in sample, recorded on a desktop browser. Your run may word things slightly differently.

Feedback draft: Your thesis is much clearer this time, but the evidence still jumps between topics. Add one source per paragraph and slow down the conclusion.

Overall

Negative

Sentence balance

0 positive · 2 negative

Most negative sentence

Add one source per paragraph and slow down the conclusion.

Most positive sentence

Feedback draft: Your thesis is much clearer this time, but the evidence still jumps between topics.

The model

The files are served from this site, versioned so a cached copy is never stale. Nothing is fetched from a third party while you use the tool.

You can switch models in the tool. Also available: Gemma 4 E2B instruction-tuned (web build) (2.0 GB). The larger language model gives more thorough answers but is a bigger download and needs a stronger device.

Limitations

• The model only knows positive and negative. It does not detect anger, urgency or politeness, and it was trained on movie reviews, so sarcasm, jargon and very short replies can be misread.
• Results describe the input you gave, not the real-world situation behind it. Review them yourself before acting.

Good uses

• Triage education and study notes before deeper review
• Prepare consistent labels for repeat workflows
• Reduce manual sorting on small batches
• Keep sensitive drafts in the browser preview