Fyn, the free document extraction tool, launches in beta!
After several months of private beta testing, Fyn is launching in open beta
While software is eating the world, a shocking amount of human labor goes into transcribing documents (e.g. think of financial documents such as invoices and receipts) into a computer-readable format. While this work of transcribing is boring and repetitive for humans, it is still a hard problem for computers.
To solve this problem, we have been hard at work developing Fyn. Fyn is a tool that eliminates boring, repetitive manual transcription of documents into structured, computer-readable formats.
Clearly, we aren't the first one to tackle this challenge, but a couple of things make us very different from other solutions.
What makes Fyn unique
Rather than hard-coding rules to recognize certain templates, Fyn reads documents the way a human does, meaning it can also extract content from templates it has never seen. This is important, because most documents have many, many different formats, meaning template-based approaches require continuous human labor to ensure they work well. Fyn is the opposite: the more examples it sees, the better it works.
Even though Fyn is still in its infancy and hasn't seen many documents yet, it's still capable of "generalizing" - that's machine learning speak for "capable of extracting content from templates it has never seen before". You can think of Fyn like a baby Superman who is already more powerful than some traditional invoice reading solutions.
What's in the Beta
- Documents you can extract: Our open beta allows you to extract the content of french, European style invoices for free
- Speed: An extraction is usually done in less than 10 seconds per document.
- Output: The extracted content is available either in a json format, or in a ubl xml format, which can directly be imported in most accounting software.
- User Interface: You can either use our dashboard or our API to upload documents and view the results.
- Pricing: All extractions are free-of-charge during the beta!
- Improving the detection quality: The beta is only trained on a tiny amount of invoices. We're working with select partners to expand our training set, so you should see further improvements of the engine in the next weeks and months.
- More languages: We'll expand the engine's language skills soon.
- More documents: We'll be working on extracting the content of receipts next - never again will you have to manually encode restaurant expenses!
- More ways to parse a document: We'll be adding the possibility to send us documents through email, which should further reduce the friction of using Fyn for personal use.
- More dashboard capabilities: We'll add the possibility to visualize the extracted results on the dashboard, and correct any mistakes or omissions in the extraction.