Machine learning based information extraction
Computer-readable media, methods, and systems are disclosed for applying machine learning mechanisms to classify and validate documents based on expense rule sets and external data validation services. Document images associated with expenses are received in connection with a reimbursable event. For each received document image data associated with the received document image is transmitted to an optical character recognition image processor that can recognize contents and associated coordinates. OCR data is received and transmitted to a text tokenizer. Tokenized text is received corresponding to expense details, and the tokenized text and coordinates are sent to a text feature generator. Text feature vectors are received and transmitted to a document classifier and a document classification received. Document fields are extracted and based thereon a document is validates and a corresponding reimbursement instruction generated.
1. One or more non-transitory computer-readable media storing computer-executable instructions that, when executed by a processor, perform a method for applying machine learning techniques to classify and validate documents based on one or more expense rule sets and one or more external data validation services, the method comprising:
receiving, by a document classification service, one or more document images associated with expenses incurred in connection with a reimbursable event;
for each received document image in the one or more document images:
transmitting image data associated with the received document image to an optical character recognition image processor, the optical character recognition image processor configured to recognize textual contents and coordinates associated with graphical and textual information contained within the received document image;
receiving optical character recognition data from the optical character recognition image processor, wherein the optical character recognition data comprises the textual contents and coordinates associated with graphical and textual information;
transmitting the optical character recognition data to a text tokenizer;
receiving tokenized text from the text tokenizer, wherein the tokenized text comprises text entities corresponding to expense details associated with the expenses incurred in connection with a reimbursable event;
transmitting the tokenized text and the coordinates associated with graphical and textual information to a text feature generator;
receiving one or more text feature vectors from the text feature generator;
transmitting the one or more text feature vectors to a document classifier;
receiving a document classification from the document classifier, wherein the document classifier employs a document classification machine learning model that is trained on previously validated documents;
extracting extracted document fields from the one or more text feature vectors based on a document extraction machine learning model and the one or more expense rule sets;
based on the extracted document fields and the document classification, validating a document based on the received document image, the received document classification, and the one or more data validation services to produce a validation result; and
based on the validation result, automatically generating a reimbursement instruction corresponding to the received document classification and the extracted document fields.
2. The non-transitory computer-readable media of claim 1 , wherein the extracted document fields comprise one or more date fields associated with the reimbursable event and wherein automatically generating a reimbursement instruction corresponding to the received document classification and the extracted document fields further comprises automatically generating a reimbursement instruction based on the one or more date fields.
3. The non-transitory computer-readable media of claim 1 , wherein the extracted document fields comprise one or more transaction location fields associated with a geographical location and wherein automatically generating a reimbursement instruction corresponding to the received document classification and the extracted document fields further comprises automatically generating a reimbursement instruction based on the one or more transaction location fields.
4. The non-transitory computer-readable media of claim 1 , wherein the method further comprises:
loading a loaded document classification machine learning model, wherein the loaded document classification machine learning model is trained based on previously approved documents.
5. The non-transitory computer-readable media of claim 4 , wherein the method further comprises:
retraining the loaded document classification machine learning model based on the reimbursement instruction.
6. The non-transitory computer-readable media of claim 1 , wherein the method further comprises:
receiving stored location information associated with the one or more document images and correlating the location information with event location information associated with the reimbursable event.
7. The non-transitory computer-readable media of claim 1 , wherein the document classifier employs a document classification machine learning model that is trained on previously invalidated documents.
8. A method for applying machine learning techniques to classify and validate documents based on one or more expense rule sets and one or more external data validation services, the method comprising:
receiving, by a document classification service, one or more document images associated with expenses incurred in connection with a reimbursable event;
for each received document image in the one or more document images:
transmitting image data associated with the received document image to an optical character recognition image processor, the optical character recognition image processor configured to recognize textual contents and coordinates associated with graphical and textual information contained within the received document image;
receiving optical character recognition data from the optical character recognition image processor, wherein the optical character recognition data comprises the textual contents and coordinates associated with graphical and textual information;
transmitting the optical character recognition data to a text tokenizer;
receiving tokenized text from the text tokenizer, wherein the tokenized text comprises text entities corresponding to expense details associated with the expenses incurred in connection with a reimbursable event;
transmitting the tokenized text and the coordinates associated with graphical and textual information to a text feature generator;
receiving one or more text feature vectors from the text feature generator;
transmitting the one or more text feature vectors to a document classifier;
receiving a document classification from the document classifier, wherein the document classifier employs a document classification machine learning model that is trained on previously validated documents;
extracting extracted document fields from the one or more text feature vectors based on a document extraction machine learning model and the one or more expense rule sets;
based on the extracted document fields and the document classification, validating a document based on the received document image, the received document classification, and the one or more data validation services to produce a validation result; and
based on the validation result, automatically generating a reimbursement instruction corresponding to the received document classification and the extracted document fields.
9. The method of claim 8 , wherein the extracted document fields comprise one or more date fields associated with the reimbursable event and wherein automatically generating a reimbursement instruction corresponding to the received document classification and the extracted document fields further comprises automatically generating a reimbursement instruction based on the one or more date fields.
10. The method of claim 8 , wherein the extracted document fields comprise one or more transaction location fields associated with a geographical location and wherein automatically generating a reimbursement instruction corresponding to the received document classification and the extracted document fields further comprises automatically generating a reimbursement instruction based on the one or more transaction location fields.
11. The method of claim 8 , further comprising:
loading a loaded document classification machine learning model, wherein the loaded document classification machine learning model is trained based on previously approved documents.
12. The method of claim 11 , further comprising:
retraining the loaded document classification machine learning model based on the reimbursement instruction.
13. The method of claim 8 , further comprising:
receiving stored location information associated with the one or more document images and correlating the location information with event location information associated with the reimbursable event.
14. The method of claim 8 , wherein the document classifier employs a document classification machine learning model that is trained on previously invalidated documents.
15. A system for applying machine learning techniques to classify and validate documents based on one or more expense rule sets and one or more external data validation services, the system comprising:
at least one processor;
and at least one non-transitory memory storing computer executable instructions that when executed by the at least one processor cause the system to carry out actions comprising:
receiving, by a document classification service, one or more document images associated with expenses incurred in connection with a reimbursable event;
for each received document image in the one or more document images:
transmitting image data associated with the received document image to an optical character recognition image processor, the optical character recognition image processor configured to recognize textual contents and coordinates associated with graphical and textual information contained within the received document image;
receiving optical character recognition data from the optical character recognition image processor, wherein the optical character recognition data comprises the textual contents and coordinates associated with graphical and textual information;
transmitting the optical character recognition data to a text tokenizer;
receiving tokenized text from the text tokenizer, wherein the tokenized text comprises text entities corresponding to expense details associated with the expenses incurred in connection with a reimbursable event;
transmitting the tokenized text and the coordinates associated with graphical and textual information to a text feature generator;
receiving one or more text feature vectors from the text feature generator;
transmitting the one or more text feature vectors to a document classifier;
receiving a document classification from the document classifier, wherein the document classifier employs a document classification machine learning model that is trained on previously validated documents;
extracting extracted document fields from the one or more text feature vectors based on a document extraction machine learning model and the one or more expense rule sets;
based on the extracted document fields and the document classification, validating a document based on the received document image, the received document classification, and the one or more data validation services to produce a validation result; and
based on the validation result, automatically generating a reimbursement instruction corresponding to the received document classification and the extracted document fields.
16. The system of claim 15 , wherein the extracted document fields comprise one or more date fields associated with the reimbursable event and wherein automatically generating a reimbursement instruction corresponding to the received document classification and the extracted document fields further comprises automatically generating a reimbursement instruction based on the one or more date fields.
17. The system of claim 15 , wherein the extracted document fields comprise one or more transaction location fields associated with a geographical location and wherein automatically generating a reimbursement instruction corresponding to the received document classification and the extracted document fields further comprises automatically generating a reimbursement instruction based on the one or more transaction location fields.
18. The system of claim 15 , wherein the actions further comprise:
loading a loaded document classification machine learning model, wherein the loaded document classification machine learning model is trained based on previously approved documents.
19. The system of claim 18 , wherein the actions further comprise:
retraining the loaded document classification machine learning model based on the reimbursement instruction.
20. The system of claim 15 , wherein the actions further comprise:
receiving stored location information associated with the one or more document images and correlating the location information with event location information associated with the reimbursable event.