People read a letter and know immediately who wants what by when. Software sees a sequence of words. Natural language processing closes this gap.
What is NLP?
Natural language processing (NLP) is the part of artificial intelligence that deals with human language. It enables software to analyse text, recognise its meaning and work with it: find names, sort messages, summarise long documents, answer questions.
NLP works with written text and, combined with speech recognition, with spoken language.
Text recognition reads, NLP understands
In document processing, two steps come together:
- Text recognition turns a scan or photo into text. It answers the question: which letters are on the page? More in our blog post on PDF text recognition.
- NLP works with this text. It answers the question: what does it mean?
NLP can only understand what text recognition has read correctly. That is why both have to be reliable.
What NLP can do
| Task | What happens | Example |
|---|---|---|
| Entity recognition | Finds names, dates, amounts, places | The IBAN and the sender in a letter |
| Classification | Assigns a text to a category | Complaint, order or cancellation |
| Summarisation | Condenses long texts | The key points of a 200-page file |
| Question answering | Answers questions about a text | “Is the proof of income missing?” |
| Text generation | Writes text from data or instructions | A draft reply to a request |
More on entity recognition in our blog post on entity extraction, more on classification in the post on PDF classification.
From rules to learning systems
NLP research began in the 1950s with first machine translations. For a long time, systems worked with hand-written rules. From the 1990s, statistical methods took over, and from the 2010s, neural networks. Today, large language models can write and answer questions in natural language.
For document processing, one thing matters above all: the results must be traceable. A system that answers questions about a file has to show where the answer comes from.
NLP in document processing
Reading and sorting
IDA reads documents, recognises their type and extracts the data a process needs. Deterministic: recognition, not guessing.
Understanding long files
With IDA Understanding, long, unstructured case files become manageable, with a summary, a table of contents and answers to your questions, for thousands of documents at once.
Reviewing and preparing
JAIDE Akte reads an entire case file, checks it for completeness and contradictions and answers the review questions, each answer with the passage it came from. The decision stays with the person.
Frequently asked questions
What does NLP stand for?
Natural language processing: the processing of human language by computers.
What is the difference between NLP and OCR?
OCR (text recognition) turns an image into text. NLP works with this text and recognises its meaning.
Is NLP the same as a large language model?
No. Large language models are one method within NLP. NLP also includes many other methods, such as entity recognition or classification.
How can I trust the results?
By checking them. Good systems show where each statement comes from. JAIDE gives the source for every answer, so people can check it before they decide.
See it with your documents
In a demo, we show you how IDA and JAIDE read and understand your documents.