DEJAN: RexBERT for E-Commerce SEO Tasks

Key Takeaways:

• RexBERT is a language model specifically trained on E-Commerce data that outperforms general-purpose models

• Potential tasks according to Dan Petrovic:

  1. Smarter Product Title & Description Audits Detects missing attributes, redundant phrasing, or keyword gaps across large catalogs

  2. Attribute Extraction at Scale

     Turns messy descriptions into structured filters (size, color, material, brand) for cleaner navigation and richer schema

  3. Semantic Internal Linking

     Powers smarter product & category linking by understanding synonyms, substitutes, and complements

  4. Duplicate Content Detection

     Flags near-duplicates across thousands of SKUs—guiding canonicalization, consolidation, or rewrites

  5. Meta Description & Snippet Simulation

     Predicts how search snippets might render, letting you A/B test before pushing changes live

  6. Category Page Relevance

     Classifies content against category intent (“men’s trail running shoes” vs “generic running shoes”) to strengthen topical depth

  7. Enhanced On-Site Search

     Improves query-to-product matching, reducing zero-result pages and increasing conversion.

"TL;DR: An encoder-only transformer (ModernBERT-style) for e-commerce applications, trained in three phases—Pre-training, Context Extension, and Decay—to power product search, attribute extraction, classification, and embeddings use cases. The model has been trained on 2.3T+ tokens along with 350B+ e-commerce-specific tokens"

Sources:

• https://www.linkedin.com/posts/seoguy_if-you-do-technical-ecommerce-seo-make-note-activity-7376016936016683008-PhzM/

• https://dejan.ai/blog/rexbert/

• https://huggingface.co/thebajajra/RexBERT-large