Custom Tabs, bulk actions, Single sign on, feedback features on API, news from our R&D Team
we caught another great release! We added some cool new features to lists and around the document name, making it easier to work with these two features overall. But we did not neglect our API as we added three new functionalities to bring more currently exclusive graphical user interface features onto the API.
There is also some news from our R&D department, enhancing the platform with image-based classification, quicker automated training intervals, and our data extraction's future.
If you have any feedback feel free to reach out
PS: If you want to always stay up to date as soon as we release new features, leave us your email here: Release Notes mailing list
🎥 Check out our release explainer video if you like 🎥
🏷️ Improved usability of Document Name
We introduced the document name a while back with our ongoing effort to make our user interfaces more user-friendly. While most users loved this decision and wished for us to double down, some power users mourned the missing document ID in lists and the new Validation Extraction interface.
So we added a little something for everyone!
The document name is now searchable, with an asterisk * for a wildcard search, and it is easier to quickly re-name a document out of any validation or detail view directly in the header. Our power users, on the other hand, can now quickly copy the document ID into the clipboard by a click on the document name in any document list or the ID in any Validation interface 🎉
🥰 Custom Tabs/Filters
More and more users work with the Parashift Platform every day, with different processes and workflows. Our document lists were pretty static in the way they were pre-filtered and ordered. This meant that, depending on workflow, users had to filter for the same documents manually repeatedly many times a day.
It is now possible to quickly save a filtered and ordered list as a new tab in your Document Overview (or other document lists) to avoid these tedious tasks. These savable tabs are already really neat but will show their full potential once we have reworked our lists with the grouping of items, overhauled filters & sorting, and more.
So stay tuned 🤙
🚚 Bulk Actions
Often a quick look at the preview of a document in a list (💡 Tip: Hover over thumbnails to get an enlarged preview) already tells a lot, or at least enough to do a proper classification or know which user should work on this specific document. But if you wanted to classify the document or assign it to a user, you had to open it in Validation and work on document after document.
To make this a lot easier, we created Bulk Actions. Mark the documents you want to classify or assign to a user, either one by one or the whole list (of up to 100 documents) and Classify/Assign them directly from the table header!
I feel like a broken record, repeating myself, but stay tuned for more bulk actions like delete or reset to follow 😉
🧠 News from our R&D Team
👩🏫 New Classification Training
We tweaked our automated classification trainer (tweaked is an understatement, we reworked it). It will now train more often and faster on new documents for classification. Overall we now need fewer classified documents to trigger a re-training.
What will stay is that you don't have to worry about manually starting trainings; we take care of that for you automatically.
🖼️ Image Classification
Document Classification has worked only with text until now, which leads to problems classifying documents like plans, x-rays or photos that typically don't have much text on them or where the text varies heavily from document to document.
So we added the image to the classification feature mix to counteract this problem with low-text document types. Benchmarks show that we now generally perform a lot better in classification across the board 😀
If you want to see this in action, go and watch our 1st April joke video on "Intelligent Meme Classification", where we used text plus image-based classification to classify memes.
💥 Learnings after only one document
We deployed an experimental 🔬 new extractor, accessible only to Parashift employees, for now, that will revolutionize automated data extraction from documents. After only one document, it produced already candidates on follow up documents with different layouts/text.
This is one of the big groundbreaking technologies that will make it to production this year; super excited!
🔄 One API to rule them all
Our ongoing effort to bring more functionality to our API (see https://docs.parashift.io) is getting three new additions this release. With these, it will be possible to influence the workflow, documents, and machine learning training from outside our graphical user interfaces.
You can now classify a document via API, which we will use as valid ground truth, so we will also train on these external calls and make classification better. The same goes for validating fields; you can now send external values for each field.
Last but not least, we allowed some document values like the name, custom fields and more to be updated (patched) via API.
All in all, we are looking forward to all the great integrations our partners and clients will implement with these new features.
If you have any questions or requests, reach out; always happy to help 👍
📄 Classify a document
New endpoint to classify a document
🟨 Feedback API Extraction
New endpoint to validate a document field
🩹 PATCH document attributes
Patch certain document attributes like name, custom fields, external id or the training flag.
Check out our official API documentation for details.
🙋 Single Sign On (SSO)
Single sign-on is now available with OpenID Connect. Use your existing identity solution directly with the Parashift Platform. Gone are the days of user logins and manually disabling users in multiple systems. For everyone interested, feel free to reach out to your account manager @Parashift.
💞 Other Improvements
- Added notification bubble to Comments tab in new Extraction Validation if comments present
- Unified tooltips and icons to delete rows/sections/line items in new Extraction Validation
- fixed wrong relationships in Page/File
- fixed color_jpeg returned instead of output_file
- new Extraction Validation
- fixed error showing field IDs instead of field names
- "not_for_validation" fields are now (correctly) readonly
- info-box/documents several smaller display bugs fixed
- fixed issue with saving fields and showing document preview on "big" (>40 pages, many line items) documents
- drastically reduced loading time for reports