Results Guide

This page explains how to read and discuss VersoVector results from the public repository.

The goal is not to publish every notebook output, but to guide a technical reader through the most relevant evidence.

Result categories

Result type	Where to inspect it	What it shows
Cleaned corpus	01_cleaning_pipeline.ipynb	Text normalization and metadata readiness
Feature pipeline	02_feature_pipeline.ipynb	Count, TF-IDF, custom features, sparse matrices
Supervised predictions	03_embeddings_supervised.ipynb	Multilabel emotional or thematic tag prediction
Similarity neighbors	04_embeddings_unsupervised.ipynb	Semantically close poems
Topics	04_embeddings_unsupervised.ipynb	Latent thematic signals
Clusters	04_embeddings_unsupervised.ipynb	Emergent poem groupings
Integrated table	05_supervised_unsupervised_integration.ipynb	Combined supervised and unsupervised view
Visual assets	06_visualizations.ipynb	Final plots and interpretation figures

Recommended outputs to render in this documentation

Good candidates:

• feature pipeline diagram;
• supervised metric summary table;
• tag distribution chart;
• topic summary table;
• cluster distribution chart;
• UMAP or t-SNE projection;
• similarity-neighbor example using safe short text;
• integrated result sample with limited columns.

Recommended image location

Use:

docs/public/images/versovector/

Example files:

docs/public/images/versovector/feature-pipeline.png
docs/public/images/versovector/supervised-metrics.png
docs/public/images/versovector/topic-summary.png
docs/public/images/versovector/cluster-projection.png
docs/public/images/versovector/integrated-results-sample.png

Then reference them:

![VersoVector cluster projection](/images/versovector/cluster-projection.png)

Suggested result narrative

When presenting results, use this structure:

Input text
  ↓
Cleaned representation
  ↓
Feature vector
  ↓
Predicted tags
  ↓
Nearest semantic neighbors
  ↓
Dominant topic
  ↓
Cluster/projection position
  ↓
Interpretation

Example interpretation pattern

The model predicts tags associated with memory, loss, and longing.

The nearest-neighbor results suggest that the text is semantically close to poems containing similar emotional vocabulary.

The topic model associates the text with terms related to time, absence, and inner conflict.

The projection places the poem near a cluster of emotionally introspective texts.

Public documentation note

Use curated outputs.

Avoid turning the documentation into a raw dataset dump. When including examples, prefer short excerpts, synthetic examples, public-domain texts, or summarized model outputs.