How does product Matching Work

The UPC, GTIN, and other identifiers on the product page are searched. The product title needs to get compared. Two product titles can never be the same in various stores for the same product. Deep learning techniques are used for identifying the recognizing the similarities, learning the differences, and creating word-level embedding to develop a system of representation for common words. It consists of teaching the system to identify different references and come up with a unique representation for each entity.

1. Identification: A product is identified using the title, description, images, specifications, attributes, etc. In some cases, the product title alone will offer a lot of information. The product matching system needs to learn to sort the product name from the specifications. For example, Samsung Galaxy Note 8 is the phone model and title offer additional information such as US version, memory size, factory unlocked refurbished, etc.

2. Extraction: This information needs to be extracted and sorted and placed in suitable categories such as memory size, phone model, version, etc. Various product matching methods might be used for helping the system classify the sets of information.

3. Comparison: Another comparison might be a similar product with more information in the form of a description which consists of memory and screen size information, tags in the title, and a specifications table. They help in adding more knowledge about the product. This helps the machine in identifying an exact product match or mismatch during next the comparison.

Learn More