Dynamically determining schema labels using a hybrid neural network encoder

A non-transitory computer-readable medium storing instructions that, when executed by at least one processor, cause a computer device to:

• identify a column within a digital dataset;
• determine whether the column comprises a header column type that includes a header label and whether the column comprises a cell column type that includes a populated column cell;
• select a neural network encoder model from a set of neural network encoder models comprising a header neural network encoder and a cell neural network encoder based on whether the column comprises the header column type and whether the column comprises the cell column type;
• generate a column vector embedding for the column utilizing the selected neural network encoder model;
• determine a schema label for the column based on the column vector embedding generated utilizing the selected neural network encoder model;
• and
• generate an updated dataset using the schema label and the column.

A system comprising:

• one or more memory devices comprising a digital dataset comprising a column, a plurality of schema labels, and a set of neural network encoder models comprising a header neural network encoder and a cell neural network encoder;
• and
• one or more server devices configured to cause the system to:
• select a neural network encoder model from the set of neural network encoder models based on whether the column comprises a header column type that includes a header label;
• and
• determine a schema label from the plurality of schema labels for the column by:
• generating a column vector embedding for the column utilizing the selected neural network encoder model;
• identifying schema vector embeddings for the plurality of schema labels, the schema vector embeddings generated utilizing the header neural network encoder;
• and
• comparing the column vector embedding to the schema vector embeddings.