Systems and methods for classification using structured and unstructured attributes

A system comprising:

• one or more processors;
• and
• one or more non-transitory computer-readable media storing computing instructions that, when executed on the one or more processors, cause the one or more processors to and perform functions comprising:
• receiving attribute data comprising text associated with an item to classify the item into one or more item categories, wherein the attribute data comprises:
• (i) a set of unstructured attribute data for the item, wherein the set of unstructured attribute data comprises one or more sentences;
• and
• (ii) a set of structured attribute data for the item;
• analyzing the set of unstructured attribute data by processing through a first set of one or more Long Short Term Memory (LSTM) layers of a first deep learning model to output obtain an unstructured semantic signature of an attribute for the item by at least embedding each word in the set of unstructured attribute data and passing a sequence of each word embedding to the LSTM layers, wherein the unstructured semantic signature of the attribute for the item is indicative of a probability that the item belongs to one or more first item categories based on the set of unstructured attribute data;
• analyzing the set of structured attribute data by processing through a first set of one or more Convolutional Neural Network (CNN) layers of a second deep learning model, to output a structured semantic signature of the item by receiving one or more attribute names in a random order and concatenating each one of the one or more attribute names as a respective string separated by a respective identifier, wherein the structured semantic signature of the item that is indicative of a likelihood that the item belongs to one or more second item categories based on the set of structured attribute data;
• generating a concatenated semantic signature of the attribute data by processing the unstructured semantic signature of the attribute for the item and the structured semantic signature of the item through one or more layers selected from a second set of one or more CNN layers;
• processing the concatenated semantic signature through a first set of bidirectional LSTM layers to obtain an item semantic signature that is indicative of the likelihood that the item belongs to the one or more item categories based on both the set of unstructured attribute data and the set of structured attribute data;
• and
• classifying the item in the one or more item categories in relation to the item semantic signature.