VABSA

Values and Attributes Based Sentiment Analysis

This page showcases the results generated by our custom-built VABSA models, applied to 74,774 textual reviews collected from an e-commerce platform featuring Made in Italy products in the Home & Furniture and Apparel & Accessories sectors.

Our VABSA model -Value and Attributes Based Sentiment Analysis - analyzes each review to extract:

  • Sentiment (from 1 to 5)
  • Presence or absence of specific Product Attributes (PAs)
  • Presence or absence of key Consumer Values (CVs)

What are these dimensions?

Sentiment

Sentiment indicates whether a customer’s opinion is positive, neutral, or negative based on the words and tone they use.

Product Attributes (PA)

Product Attributes are the tangible features that define a product’s quality, functionality, and appeal.

Consumer Values (CV)

Consumer Values represent the benefits or psychological satisfaction that the customer derives from the product’s attributes. They reveal deeper motivations that enhance customer engagement and loyalty.

...

Example:

  • Positive sentiment:

    “The product is beautiful and well-made.”

  • Neutral sentiment:

    “The item arrived as expected.”

  • Negative sentiment:

    “The material feels cheap and poorly assembled.”

...
Quality

Well-made, durable, and built to last. Reflects craftsmanship and material strength.

...
SUSTAINABILITY CUES

Signals eco-conscious design—recyclable, long-lasting, or repairable.

...
ECO-LABELS & PACKAGING

Official certifications or recyclable, minimal packaging that supports eco claims.

...
COUNTRY OF ORIGIN

Emphasizes authenticity, tradition, and support for local craftsmanship.

...
PRICE

Reflects value for money—whether affordable, fair, or premium for quality.

...
SERVICE (WARRANTY & AFTERSALES)

Covers delivery, customer support, and warranty reliability.

...
GREEN VALUE

Appreciation for eco-friendly, low-impact, and sustainable products.

...
FUNCTIONAL VALUE

Focuses on utility, performance, and how well the product serves its purpose.

...
SOCIAL VALUE

Enhances social image, impresses others, or suits gifting.

...
EMOTIONAL VALUE

Evokes personal satisfaction, joy, or pride in ownership.

...
CONDITIONAL VALUE

Adds meaning during specific occasions like holidays or seasonal events.

...
EPISTEMIC VALUE

Sparks curiosity, discovery, or interest—often through uniqueness or cultural depth.

All visualizations on this page — including sentiment distribution, PA/CV frequencies, Jaccard similarity heatmap, word clouds, and category-specific analyses — are entirely based on outputs from our model. These insights reveal how different product features and values relate to customer sentiment and language.
Analysis Chart

    This plot shows the distribution of sentiment scores assigned by our model to all text reviews. Sentiment ranges from 1 (very negative) to 5 (very positive), with neutral score of 3. The blue curve represents a smoothed trend of the distribution.

    Our model identifies which Consumer Values (CV) and Product Attributes (PA) are present in each review and labels them accordingly. The two plots show the frequency counts of PAs and CVs found in the dataset. By clicking a single bar you can drill down into it.

    Our model identifies which Consumer Values (CV) and Product Attributes (PA) are present in each review and labels them accordingly. The two plots show the frequency counts of PAs and CVs found in the dataset. By clicking a single bar you can drill down into it.

    Hereafter some visualization of the results, stratified by the section from which data comes from, are reported. This pie chart shows the distribution of reviews across different Amazon sections. By clicking the slice you can drill down into the section.

    In this heatmap each cell represents how often a PA and a CV appear together in the same review, relative to their total occurrences. Brighter colors indicate stronger associations, helping to highlight which product features are most linked to specific consumer values. The similarity between all Product Attributes (PAs) and Consumer Values (CVs) have been computed usign the Jaccard similarity index.