Automatic keyword extraction with machine learning for products filtering

Filters are of great importance for online stores, the more closely the demand can be filtered by the consumer, the more likely he is to come across a product that interests him. Adding filters to products is often a time-consuming and tedious process, so store employees may not spend enough time on it or simply may not have it. In many stores the products are added quickly and there are no filters. My customer with a custom online store created by me with over 50,000 items is needed to add more filters so that users could better refine their searches and find products that interested them more easily. For this purpose, I developed with the help of machine learning keyword extraction from product descriptions in Bulgarian. These keywords were later used to automatically create filters. Site administrators have the option to remove filters from appearing if the filter does not seem appropriate to them or is duplicated by another. It took 15 minutes to generate the filters for 50,000 products. It would take the site staff an awful lot of time to do it by hand.