Tapping item-to-items relationships.

Using proprietary algorithms, our platform is able to map the complex relationships between items. After extracting deep information from an item’s descriptive text with natural language processing, or from images using advanced convolution neural networks, we can then create a rich representation of its associations to other products.

Combined with our proprietary Semantic Graph Embedding, our Deep Content Extraction allows our software to recommend items of all kinds—even items no one has interacted with yet. This is because our models are able to understand the product type or genre by automatically extracting information and text from photos, descriptions, or reviews.