Web page transformer for structure information extraction

A processing system configured for structured information extraction from electronic documents, the processing system comprising:

• memory configured to store data associated with the structured information extraction;
• and
• one or more processors operatively coupled to the memory, the one or more processors being configured to:
• obtain, at an input layer of a model, information associated with a structured electronic document, the information including (i) one or more field tokens each representing a respective field to be extracted from the structured electronic document, (ii) a set of structured document type (“S type”) tokens associated with a document object model (“DOM”) of the structured electronic document, and (iii) a set of text tokens from a text sequence in the structured electronic document;
• connect, by an encoder of the model, the one or more field tokens, the set of S type tokens and the set of text tokens according to a set of different attention patterns, the set of different attention patterns including (i) an S type-to-S type attention that models relations among the S type tokens via graph attention, (ii) an S type-to-text attention that bridges each respective S type token in the set with a corresponding text token from the set of text tokens, (iii) text-to-S type attention that propagates information from the set of S type tokens to the set of text tokens, and (iv) text-to-text attention;
• generate, by the encoder, an overall token representation based on the set of different attention patterns;
• extract, by an output layer of the model, a final text span for the each of the respective fields from the set of text tokens;
• and
• store, in the memory, the extracted final text span for each of the respective fields in memory associated with the computing system, the extracted final text span for each of the respective fields being configured for production in response to a request.