AI-augmented auditing platform including techniques for automated document processing
Systems and methods for automated document processing for use in AI-augmented auditing platforms are provided. A system for determining the composition of document bundles extracts substantive content information and metadata information from a document bundle and generates, based on the extracted information regarding a composition of the document bundle. A system for validating signatures in documents extracts data representing a spatial location for respective signatures and generates a confidence level for respective signatures, and determines, based on location and confidence level, whether signature criteria are met. A system for extracting information from documents applies a set of data conversion processing steps to a plurality received documents to generate structured data, and then applies a set of knowledge-based modeling processing steps to the structured data to generating output data extracted from the plurality of electronic documents.
1 . A system for validating signatures in documents, the system comprising one or more processors configured to cause the system to:
receive an electronic document comprising one or more signatures;
process the electronic document to generate a set of signature criteria for the received electronic document:
apply one or more signature-extraction models to the electronic document to generate, for each of the one or more signatures in the electronic document, data representing a spatial location for the respective signature and a confidence level for the respective signature; and
determine, based on the data representing the spatial location and the confidence level, whether the electronic document satisfies the set of signature criteria.
2 . The system of claim 1 , wherein the one or more signature-extraction models comprise a first signature-extraction model configured to recognize signatures regardless of spatial location.
3 . The system of claim 1 , wherein the one or more signature-extraction models comprise a second signature-extraction model configured to recognize signatures based on in-document spatial location.
4 . The system of claim 3 , wherein applying the second signature- extraction model comprises:
determining a predicted spatial location within the electronic document based on one or more of a structure, format, and type of the electronic document; and
extracting a signature from the predicted spatial location.
5 . The system of claim 1 , wherein determining whether the electronic document satisfies the set of signature criteria comprises determining whether a signature appears in the electronic document at a required spatial location.
6 . The system of claim 1 , wherein determining whether the electronic document satisfies the set of signature criteria comprises determining whether the confidence level exceeds a predefined threshold.
7 . The system of claim 1 , wherein determining whether the electronic document satisfies the set of signature criteria comprises determining whether a signature appears in the electronic document within a required spatial proximity to a component extracted from the document.
8 . The system of claim 1 , wherein determining whether the electronic document satisfies the set of signature criteria comprises generating an association score indicting a level of association between a signature extracted from the electronic document and signature-context data generated based on the electronic document.
9 . The system of claim 1 , wherein the system is configured to determine the set of signature criteria based at least in part on context data, wherein the context data indicates one or more of: document type, document structure, and document format.
10 . The system of claim 1 , wherein the system is configured to determine the set of signature criteria based at least in part on the one or more signatures detected in the document.
11 . A non-transitory computer-readable storage medium storing instructions for validating signatures in documents, the instructions configured to be executed by a one or more processors of a system to cause the system to:
receive an electronic document comprising one or more signatures;
process the electronic document to generate a set of signature criteria for the received electronic document:
apply one or more signature-extraction models to the electronic document to generate, for each of the one or more signatures in the electronic document, data representing a spatial location for the respective signature and a confidence level for the respective signature; and
determine, based on the data representing the spatial location and the confidence level, whether the electronic document satisfies the set of signature criteria.
12 . A method for validating signatures in documents, wherein the method is performed by a system comprising one or more processors, the method comprising:
receiving an electronic document comprising one or more signatures;
processing the electronic document to generate a set of signature criteria for the received electronic document;
applying one or more signature-extraction models to the electronic document to generate, for each of the one or more signatures in the electronic document, data representing a spatial location for the respective signature and a confidence level for the respective signature; and
determining, based on the data representing the spatial location and the confidence level, whether the electronic document satisfies the set of signature criteria.
13 . A system for extracting information from documents, the system comprising one or more processors configured to cause the system to:
receive a first data set comprising a first plurality of electronic documents;
apply a set of data conversion processing steps to the first plurality of electronic documents to generate a first processed data set comprising first structured data generated based on the first plurality of electronic documents, wherein applying the set of data conversion processing steps comprises applying one or more deep-learning-based optical character recognition (OCR) models;
train a knowledge-based deep learning model trained based on the first structured data and a first plurality of data labels indicated by one or more user inputs;
receive a second data set comprising a second plurality of electronic documents;
apply the set of data conversion processing steps to the second plurality of electronic documents to generate a second processed data set comprising second structured data generated based on the second plurality of electronic documents, wherein applying set of data conversion processing steps comprises applying one or more deep-learning-based optical character recognition (OCR) models; and
apply a set of knowledge-based modeling processing steps to the second structured data, wherein applying the set of knowledge-based modeling processing steps comprises:
applying the knowledge-based deep learning model to the second structured data; and
generating output data extracted from the second plurality of electronic documents by the deep learning model.
14 . The system of claim 13 , wherein applying the set of data conversion processing steps comprises, before applying the one or more deep-learning-based OCR models, applying an automated orientation correction processing step.
15 . The system of claim 13 , wherein applying the set of data conversion processing steps comprises, before applying the one or more deep-learning-based OCR models, applying a denoising function.
16 . The system of claim 13 , wherein applying the one or more deep-learning-based OCR models comprises:
applying a text-detection model; and
applying a text-recognition model.
17 . The system of claim 13 , wherein applying the set of data conversion processing steps comprises, after applying the one or more deep-learning-based OCR models, generating the structured data based on an image-level feature engineering step.
18 . The system of claim 13 , wherein applying the set of data conversion processing steps comprises applying a post-processing method that uses morphology to parse structural relationships amongst words.
19 . The system of claim 13 , wherein applying the set of knowledge-based modeling processing steps comprises, before receiving the user input indicating the plurality of data labels, generating the structured data based on one or more feature engineering processing steps.
20 . The system of claim 19 , wherein the one or more feature engineering processing steps comprise predicting word groups based on morphology.
21 . The system of claim 13 , wherein applying the set of knowledge-based modeling processing steps comprises applying a model trained based on user used for user-defined feature engineering.
22 . The system of claim 13 , wherein applying the set of knowledge-based modeling processing steps comprises applying fuzzy matching, wherein the system is configured to consider a partial match sufficient for labeling purposes, to automatically label documents on a word-by-word basis.
23 . The system of claim 13 , wherein applying the set of knowledge-based modeling processing steps comprises automatically correcting one or more text-recognition errors during a training process.
24 . The system of claim 13 , wherein the knowledge-based deep learning model comprises a loss function that is configured to accelerate convergence of the knowledge-based deep learning model.
25 . The system of claim 13 , wherein the knowledge-based deep learning model comprises one or more layers using natural language processing (NLP) embedding such that the model learns both content information and related location information.
26 . The system of claim 13 , wherein the knowledge-based deep learning model is trained using an adaptive feeding method.
27 . The system of claim 13 , wherein the knowledge-based deep learning model comprises an input layer that applies merged embedding.
28 . The system of claim 13 , wherein the knowledge-based deep learning model comprises an input layer that is configured for variant batch sizes.
29 . The system of claim 13 , wherein the knowledge-based deep learning model comprises an input layer that applies a sliding window.
30 . The system of claim 13 , wherein the knowledge-based deep learning model comprises one or more fully-dense layers disposed between an input layer and a prediction layer.
31 . The system of claim 13 , wherein the knowledge-based deep learning model comprises a prediction layer that generates one or more metrics for presentation to a user.
32 . A non-transitory computer-readable storage medium storing instructions for extracting information from documents, the instructions configured to be executed by one or more processors of a system to cause the system to:
receive a first data set comprising a first plurality of electronic documents;
apply a set of data conversion processing steps to the first plurality of electronic documents to generate a first processed data set comprising first structured data generated based on the first plurality of electronic documents, wherein applying the set of data conversion processing steps comprises applying one or more deep-learning-based optical character recognition (OCR) models;
train a knowledge-based deep learning model trained based on the first structured data and a first plurality of data labels indicated by one or more user inputs;
receive a second data set comprising a second plurality of electronic documents;
apply the set of data conversion processing steps to the second plurality of electronic documents to generate a second processed data set comprising second structured data generated based on the second plurality of electronic documents, wherein applying set of data conversion processing steps comprises applying one or more deep-learning-based optical character recognition (OCR) models; and
apply a set of knowledge-based modeling processing steps to the second structured data, wherein applying the set of knowledge-based modeling processing steps comprises:
applying the knowledge-based deep learning model to the second structured data; and
generating output data extracted from the second plurality of electronic documents by the deep learning model.
33 . A method for extracting information from documents, wherein the method is executed by a system comprising one or more processors, the method comprising:
receiving a first data set comprising a first plurality of electronic documents;
applying a set of data conversion processing steps to the first plurality of electronic documents to generate a first processed data set comprising first structured data generated based on the first plurality of electronic documents, wherein applying set of data conversion processing steps comprises applying one or more deep-learning-based optical character recognition (OCR) models;
training a knowledge-based deep learning model trained based on the first structured data and a first plurality of data labels indicated by one or more user inputs;
receiving a second data set comprising a second plurality of electronic documents;
applying the set of data conversion processing steps to the second plurality of electronic documents to generate a second processed data set comprising second structured data generated based on the second plurality of electronic documents, wherein applying set of data conversion processing steps comprises applying one or more deep-learning-based optical character recognition (OCR) models; and
applying a set of knowledge-based modeling processing steps to the second structured data, wherein applying the set of knowledge-based modeling processing steps comprises:
applying the knowledge-based deep learning model to the second structured data; and
generating output data extracted from the second plurality of electronic documents by the deep learning model.