Layout-agnostic clustering-based classification of document keys and values

A computer-implemented method comprising:

• obtaining an electronic image comprising a visual representation of a form;
• generating, using a machine learning (ML) model, a plurality of feature vectors corresponding to a plurality of text elements of the form represented within the electronic image;
• clustering the plurality of feature vectors into at least two clusters;
• determining that a first cluster of the at least two clusters includes feature vectors corresponding to either keys of the form or values of the form;
• and
• storing at least the values of the form in a storage location.

A system comprising:

• a storage service implemented by a first one or more electronic devices within a provider network;
• and
• a document processing service implemented by a second one or more electronic devices within the provider network, the document processing service including instructions that upon execution cause the document processing service to:
• obtain an electronic image comprising a visual representation of a form from the storage service;
• generate, using a machine learning (ML) model, a plurality of feature vectors corresponding to a plurality of text elements of the form represented within the electronic image;
• cluster the plurality of feature vectors into at least two clusters;
• determine that a first cluster of the at least two clusters includes feature vectors corresponding to either keys of the form or values of the form;
• and
• store at least the values of the form at a storage location of the storage service.