PERCEPTOR, our AI to extract information from any document, is a unique blend of natural language and image processing techniques. It's ready to go out of the box and can easily be fine-tuned with about 50 documents.
People deal with documents every day. But documents like invoices or delivery bills are not just text: they have structure. And structure carries meaning. Unfortunately, computers are just bad at understanding contextual structure and thus automatically processing documents is really, really hard.
PERCEPTOR is here to change document processing: It looks at a document and understands not just the words, but also the contextual structure. So it can extract any information from any document, just like humans do. And there's more! PERCEPTOR understands various languages for data extraction and analysis. You got an Italian document but only speak German? PERCEPTOR has got you covered! To explain PERCEPTORS decisions, we can also show where PERCEPTOR searches for information within a document. How does PERCEPTOR work? PERCEPTOR is more than text processing. Combining state of the art natural language processing with image recognition, PERCEPTOR learned how humans structure information. It looks at the text, but also where text is located. And what text is next, above, and below. Just like humans do. And if PERCEPTOR cannot find a solution, we can adjust it to your needs with just 50 examples from your daily documents.
PERCEPTOR is a new take on document processing. No OCR, no fixed masked, no hard-coded rules. Just an AI looking at a scanned document. Currently, PERCEPTOR helps companies to automatically process all kinds of documents, but we want to take PERCEPTOR beyond mere automation: PERCEPTOR could be an assistant for everyone dealing with documents every day. Its multimodal and multilingual capabilities can help with foreign languages or by creating more inclusive applications.