Systems and methods for text simplification with document-level context

A system for training a text simplification model, the system comprising:

• a memory that stores a text simplification model and a plurality of processor-executable instructions;
• a communication interface that receives a training dataset including a first set of original textual samples and original revision histories of original textual samples and a second set of simplified textual samples and simplified revision histories of simplified textual samples, wherein each original revision history of a corresponding original textual sample includes a sequence of revisions to the original textual sample at a first set of times, and each simplified revision history of a corresponding simplified textual sample includes a sequence of revisions to the simplified textual sample at a second set of times;
• and
• one or more hardware processors that read and execute the plurality of processor-executable instructions from the memory to perform operations comprising:
• identifying a training pair including an original textual sample and the corresponding original revision history from the first set and a counterpart simplified textual sample and the corresponding simplified revision history from the second set;
• generating, via a neural network-based alignment model implemented on one or more hardware processors, an alignment label reflecting a first revision in the corresponding original revision history and a second revision in the corresponding simplified revision history belong to a same type of revision;
• generating, via a neural network-based classification model implemented on one or more hardware processors, a revision category label reflecting a category of the same time of revision;
• storing an updated training dataset including the training pair;
• and
• training a neural network-based text simplification model based on the updated training dataset.

A non-transitory machine-readable medium comprising a plurality of machine-executable instructions which, when executed by one or more processors, are adapted to cause the one or more processors to perform operations comprising:

• receiving, via a data interface, a training dataset including a first set of original textual samples and original revision histories of original textual samples and a second set of simplified textual samples and simplified revision histories of simplified textual samples, wherein each original revision history of a corresponding original textual sample includes a sequence of revisions to the original textual sample at a first set of times, and each simplified revision history of a corresponding simplified textual sample includes a sequence of revisions to the simplified textual sample at a second set of times;
• identifying a training pair including an original textual sample and the corresponding original revision history from the first set and a counterpart simplified textual sample and the corresponding simplified revision history from the second set;
• generating, via a neural network-based alignment model implemented on one or more hardware processors, an alignment label reflecting a first revision in the corresponding original revision history and a second revision in the corresponding simplified revision history belong to a same type of revision;
• generating, via a neural network-based classification model implemented on one or more hardware processors, a revision category label reflecting a category of the same time of revision;
• storing an updated training dataset including the training pair;
• and
• training a neural network-based text simplification model based on the updated training dataset.