Generalized enterprise catalog classification for shorthand item descriptors

A method for classifying shorthand item descriptors in accordance with an enterprise catalog, the method comprising:

• receiving a shorthand descriptor of an item;
• inputting the shorthand descriptor into an first model, the first model being a probabilistic model configured to predict characters for insertion between adjacent characters in input text;
• receiving, as output from the first model, a normalized descriptor of the item corresponding to the shorthand descriptor, wherein the first model determines the normalized descriptor by:
• receiving a plurality of candidate item descriptors as output from the first model, each of which correspond to the shorthand descriptor, the plurality of candidate item descriptors including adjacent characters of the shorthand descriptor and one or more additional characters inserted between the adjacent characters;
• identifying a candidate item descriptor with a highest probability of matching the shorthand descriptor relative to the plurality of candidate item descriptors;
• and
• assigning the candidate item descriptor with the highest probability of matching the shorthand descriptor as the normalized descriptor;
• determining one or more categories corresponding to the normalized descriptor;
• inputting the normalized descriptor and the one or more categories into a second model,
• wherein the second model is trained on data corresponding to the enterprise catalog using a supervised learning process;
• and
• receiving as output from the second model an identification of an item included in the enterprise catalog corresponding to the normalized descriptor.

A system for classifying shorthand descriptors in accordance with an enterprise catalog, the system comprising:

• an item descriptor normalization module for receiving a shorthand descriptor of an item;
• a character infilling model for taking the shorthand descriptor as input, and outputting a normalized descriptor of the item, the character infilling model being a probabilistic model configured to predict characters for insertion between adjacent characters in input text by:
• receiving a plurality of candidate item descriptors as output from the first model, each of which correspond to the shorthand descriptor, the plurality of candidate item descriptors including adjacent characters of the shorthand descriptor and one or more additional characters inserted between the adjacent characters;
• identifying a candidate item descriptor with a highest probability of matching the shorthand descriptor relative to the plurality of candidate item descriptors;
• and
• assigning the candidate item descriptor with the highest probability of matching the shorthand descriptor as the normalized descriptor, wherein the item descriptor normalization module is further configured to determine one or more categories corresponding to the normalized descriptor;
• and
• a catalog matching model for taking as input the normalized item descriptor and the one or more categories, and outputting an identification of an item included in the enterprise catalog corresponding to the normalized descriptor, wherein the second model is trained on data corresponding to the enterprise catalog using a supervised learning process.