The question posed by Capable-Pie7188 is a good one, but it's aimed at the wrong starting point. You don't cluster your way into categories. You define the categories you need, then use clustering to sort products into them. The distinction matters because it changes the entire workflow from a fuzzy exploratory exercise into a disciplined data operation.
For a furniture and decor business, your product titles and descriptions are already rich with categorical signals. Words like "sofa," "table," "lamp," and "rug" are not ambiguous. Dimensions and weight give you physical constraints that text alone cannot. The practical move is to begin with a rule-based pass that assigns products to broad buckets based on these obvious keywords and numeric thresholds. That gets you 70 to 80 percent of the way there with zero machine learning. Then you use clustering on the remaining ambiguous items, the ones where a "chair" might be an accent piece or an office seat, to refine the groupings. This hybrid approach is faster, more transparent, and easier to audit than a pure unsupervised model.
The advanced steps you mention later, things like attribute extraction or duplicate detection, only work well after you have stable categories. Once you have clean buckets, you can train a classifier to assign new products automatically. You can also use the clusters to surface inconsistencies, like a product whose description says "wood" but whose dimensions suggest particleboard. That is where the real value lives. Not in the clustering itself, but in the structure it reveals.
So here is the concrete advice. Start with a simple keyword and dimension-based taxonomy. Test it against a sample of 100 products you have labeled by hand. Then apply clustering only to the leftovers. Measure how often the cluster assignments match your manual labels. Iterate until you hit a threshold you trust. That process is not glamorous, but it is how you turn messy product data into a reliable system. Skip the search for a perfect algorithm. Build the framework first, and the categories will follow.