Document labeling
Context
How to label documents of various types and determine their usability?
A bank specializing in automobile financing for individuals needs to improve its process for processing loan applications. Each reseller sends the financing request documents in files without description. Part of the recognition of documents is done by an offshore service and knowing if the files are complete takes 3 to 4 days.
The mission consists of a double labeling of documents. On the one hand, knowing what type is a document among the 25 possible types; on the other hand, correctly label the usable and unusable documents, in order in the latter case, to immediately send a request for confirmation to the reseller.lui permettant de renvoyer le dossier.
Example of an exploitable document
Example of non-exploitable document
Solution
Scenario used
1
OCR Worker
Recovers text from a scanned image (pdf of photos)
2
ETLWorker
Cleans and formats data and categorizes documents
3
Prediction Worker
Determines the operability of documents
4
API Worker
Sends the files to the validation service
Return on investment
Cleaning, standardization, control and consolidation of information.
Automatic double labeling of documents
3 days saved on the financing request