Scalable neural tensor network with multi-aspect feature interactions

A processing computer comprising:

• a processor;
• and
• a computer-readable medium coupled to the processor, the computer-readable medium comprising code executable by the processor for implementing a method comprising:
• determining a set of regions for each of a first plurality of images of a first item type, a second plurality of images of a second item type, and a third plurality of images of a third item type;
• for each region in each set of regions of the first plurality of images, the second plurality of images, and the third plurality of images, generating a vector representing the region;
• generating a plurality of aggregated messages using the vectors corresponding to combinations of images of different types of items, the images being from the first, second, and third plurality of images;
• generating unified embeddings for the images in the first, second, and third plurality of images, respectively, using aggregated messages in the plurality of aggregated messages;
• and
• creating matching scores associated with combinations of the images of the first item type, the second item type, and the third item type using the unified embeddings and a machine learning model.