OCR + NLP for contract digitisation — 90% straight-through.

Oil & Gas / OCR / NLP / Document Intelligence

The project uses OCR and NLP to digitise documents and streamline document-based workflows in the Oil & Gas domain. The solution processes image/PDF documents and converts their contents into structured information — helping users search contracts by keyword, collect document data, identify relationships between documents, extract relevant phrases and clauses, and read and recognise entities from ID documents.

Business requirements & challenges

Business requirements & challenges

The domain is document-heavy — contracts, legal files and ID documents — and needed to become searchable, connected and structured without manual review.

Key requirements

• Keyword-based contract search across a large corpus
• Collect document data and make it available for further processing
• Identify clues indicating that one document is related to another
• Extract key phrases and clauses relevant to the required analysis from full contracts
• Automatically read and recognise ID documents

Solution

Solution

The solution combines web scraping, document preprocessing, OCR, document classification, NLP and Named Entity Recognition. A scraping tool harvests legal documents to build datasets; images and PDFs are cleaned and annotated for NER; and different NER models are trained per ID document type.

• Legal document scraping — collect documents from source websites and create a dataset
• Image/PDF cleaning and annotation for Named Entity Recognition
• NER model training on the annotated data
• Different NER models trained per ID document type for recognition accuracy
• Converts unstructured document content into structured entity-value information (e.g. Contract ID, Date, Grantor, Grantee)

Stack

Technologies used

How it works

Process workflow

1. 01Input — documents provided as images or PDFs (contracts, legal or ID documents)
2. 02Image pre-processing — noise removal, de-skewing and watermark removal
3. 03Image-to-text — document classification, layout extraction, OCR and barcode processing
4. 04Text processing — text pre-processing, text classification and entity extraction
5. 05Output — structured entities (entity → value), not just raw OCR text

What's new

Key innovations

Business benefits

Business benefits

A large proportion of the document-processing workflow runs automatically, with high accuracy and a major reduction in manual effort.

At a glance

Key highlights