Document News

articleReducto Blog09.14.26

How Finch Legal Uses Reducto to Help Paralegals Manage 5x More Personal Injury Cases Reducto

ANALYSISUNCLEAR

Finch Legal: citations for checking the original document

My take — OCR is part of how Finch Legal got its paralegals to handle 5 times the usual caseload. Normally, paralegals pick up a case they haven't touched in a while and they need to call the client, but first they need to go through medical records or email or insurance paper to figure out what all happened since they last looked. Finch says that catch-up took about a half hour before they could do the phone call. Now Finch starts working through the paperwork as it arrives. Reducto, an OCR vendor, reads every document and keeps track of where the information came from. Finch then has a system to show what changed, flag a missing signature, get the paralegal's attention. When the paralegal needs to look into something deeper, they have a citation that jumps them back to the spot in the actual original document. This is important because if fact-checking takes as long as it would be to go through it manually, it doesn't save any time. The citation is what lets them have a review, understand what it says, jump back and quickly fact-check, and save all that time. In most real-world product workflows, this is pretty necessary.

X@VikParuchuri09.09.26

New API Option Generates PDFs Passing Automated Accessibility Checks

PRODUCT/API UPDATEMATERIAL

Datalab: readable text and document structure for screen readers

My take — Datalab has an API that adds readable text and document structure to PDFs for screen readers. Let's look at the calculus example. We can see from the original text that there's all kinds of structure there. There's integral calculus on the top. There's prove that with a number next to it, and underneath that there is an equation, and we can see that it's indented. And so there's all kinds of visual information that just pure text doesn't represent. And we can see that when we look at this text layer. It's very garbled. It's not something that you would want to read and make any sense of. So instead, if you use the Datalab API, you can see that the structure is extracted. We see text and formula and text and text and formula and text, all of these in different sections that can be nicely parsed and rendered depending on the type, formulas get rendered as formulas, and then text can get rendered that way. I would still have a human in the loop check this, but this is really cool because PDFs often are not very accessible to people who are blind or other things. And so this is a step to helping that, and hopefully it just reduces the cost of getting really good translations and really good accessible ways of reading existing PDF documents.

videoLandingAI09.08.26

LandingAI Introduces Split API to Separate Bundled Documents into Typed Sub-Documents

PRODUCT/API UPDATENEW

LandingAI Split: separating bundled documents

My take — LandingAI Split takes a parsed PDF and the descriptions of document types that you want. It'll return each document's type, pages, and text, and so intake systems can send it to the right type of extraction. I tried two sample packets with this, a lease with its attachments and a medical chart from a hospital stay. There's 18 pages and it comes as one file, as leases often do, but there's actually 8 different documents in there. The 6-page lease, house rules, security deposit policy, concession acknowledgement, and the bed bug, mold, pet, and satellite addendum. LandingAI split this into all the different documents just as you would expect. Some of them were 6 pages, the pet addendum was 3, and it worked perfect. The second was a sample chart. It includes a registration, a consultation, an operation, progress notes, and all kinds of other medical things. Here we want to split it by report. So the lab report can go to a lab results extractor and an operative report can go to a separate extraction. We can't really split it based on condition because the diabetes and heart disease may be on the same page and scattered across pages. This worked really well except for one page. This page is explicitly marked Medical Report Continued, and so I would have expected it to get this split correct. In both cases, this is very useful. You often have these mixed jumbles of reports, especially when a customer or a patient is filling out forms where they might have 5 or 6 different forms and they just get scanned together. And breaking them up really makes things a lot easier.

X@datalabto09.04.26

Chandra Open-Source Tool Accelerates STEM Braille Transcription Workflow

ANALYSISNO RUNTIME CHANGE

articleDatalab Blog09.04.26

Davis County School District uses Chandra OCR to streamline STEM braille transcription

ANALYSISNO RUNTIME CHANGE

OCR in the Schools: STEM Materials in Braille

My take — A teacher in Davis County is using Chandra's OCR to help get math and science students materials in braille. He's a certified braille transcriber, and this lets him get teaching materials ready for students where state transcription services don't have them ready yet. Some students arrive during the school year. Some curriculum gets written after the final deadline for sending it to the state. So Ryan Hunsaker has to produce those himself. He'd already tried OCR, but it left him with equations missing and paragraphs out of order. Chandra produces Markdown and LaTeX that he sends through Pandoc into his braille tools. He says checking OCR now takes 5 or 10 minutes. I like that he measured the time he spends fixing the output. A model can get most of a page right and still leave you with lots of cleanup to the point where you haven't saved much work. Measuring the time from having model outputs ready to having something useful enough to send out to people is a really important thing to look at. This is a great example, and I love that this is helping. Datalab case study: https://www.datalab.to/blog/davis-county-datalab-casestudy

modelHugging Face09.04.26

tencent/EVIE family released

MODEL UPDATEMATERIAL

EVIE: Search document page images

My take — Tencent released EVIE, a family of open-weight models for searching document page images. You index the pages, then give it a question, and it ranks the pages that match. These matches can include charts and diagrams, the things that don't get represented very well in text. Each page is turned into multiple vectors (late interaction), so different parts of the image can contribute to the match, whether it's part of the text or the diagram. I ran it on a variety of documents that I had and gave it questions like finding the CDC fruit and vegetable chart, NASA's process diagram, and the loan estimate's cash-to-close page. The intended page was ranked first for all three. This wasn't a widespread benchmark, and I'd like to find where this fails, but in a few realistic cases it did work. If you give this a shot, I'd love to hear how it works for you in a real production workflow.

modelHugging Face09.04.26

Nutrient releases form-field-v1-state, an open-weight form field classifier

MODEL DROPNEW

Nutrient's field-state model for whether field is populated, signed, checked or not

My take — Nutrient released a state model that predicts whether a field or checkbox is blank or filled. Normally you would use a field detector, take each detected text box, checkbox, or signature, and send that small rectangle to the state model for the empty-versus-filled check. This is difficult but critical. A signed legal document and an unsigned legal document should be treated very differently. I ran it on a blank text field, which scored 0.007 filled; a typed text field, which scored 0.892; a checked box, which scored only 0.416; and a signature field with some lines in it, which scored essentially 1.0. Overall it seems pretty good. If I were to use this, I would do more checking around checkboxes to see what is needed to handle them correctly. This is not a benchmark, just a quick second-step check.

releaserun-llama/ParseBench09.03.26

ParseBench v1.0.0 released with extensive document parsing pipelines and PyPI packaging

SOFTWARE RELEASEMATERIAL

articlearXiv09.03.26

STAIR (STructure Aware Information Retriever): A novel dataset and LLM based retriever for document structure augmentation

research papernew

STAIR: Find document sections using a table of contents

My take — There's a new paper about training an LLM to find the right section of a document using its table of contents. The paper shows an example with a question about plurality voting. The table of contents points to an electoral systems section of a politics textbook. They started with books that already have a table of contents, extracted those, and mapped the sections to the text. They generate training questions from the books and fine-tune Mistral-7B to select the matching section. When you search, you put your question and the complete table of contents into the model. It gives you a section identifier, and your application can then go fetch that section's text. I wasn't able to find a public repository for the trained model that was linked in the paper, so that might be coming soon. But this idea of using the table of contents, the kind of manually curated structure, to find sections of documents seems really important. There's a lot of PDFs that start with a table of contents or some sort of header. I wonder if you could use a similar approach to generate a table of contents based on all of the headers. Models like Chandra identify which ones are headers, and you could use that to create the section mapping needed for this work. It would be interesting to see that and see if that helps retrieval.

articleLlamaIndex Blog09.03.26

LlamaIndex releases Turbo beta extraction tier for low-latency document processing

PRODUCT/API UPDATENEW

Turbo delivered the claimed speed in my small tests

My take — LlamaIndex added a Turbo mode in beta for much lower latency. This is super important when you have something users wait on, like uploading a document to answer questions, do retrieval, or create a report. I compared Turbo and Cost Effective mode on the same examples from my document lab. On a JPMorgan balance sheet, Turbo finished in 10.5 seconds versus 41.2 seconds, about 4 times faster. On a Vanguard table, Turbo took 9 seconds instead of 30 seconds, about 3.3 times faster. Both returned identical correct values on both documents, so there was no accuracy trade-off on either of them. Turbo is 35 credits per page, where Cost Effective is 8 credits per page. Turbo is fast, expensive, and great for regular documents. Cost Effective is slow, cheap, and great for regular documents. Agentic and Agentic Plus are slow, expensive, but very accurate on really hard documents. Overall, Turbo seems like a great fit when you need fast extraction while users are waiting. In these tiny experiments, Turbo delivered the claimed speed without the extraction quality trade-off. It wasn't a full independent benchmark, just a few tests.

articleLlamaIndex Blog09.02.26

LlamaIndex and Kaggle launch ExtractBench document extraction leaderboard

BENCHMARKNEW

ExtractBench seems like a really practical dataset

My take — LlamaIndex and Kaggle just released ExtractBench, which is a benchmark for getting structured data out of enterprise-type documents. It has about 5,000 pages in it, 370 documents, and it's across eight different domains and lots of different categories. The input is a full document with a user-defined JSON schema. Then the output is all of the values that it should extract. You're trying to get structured data that you can work with out of these documents. For the extracted values, it also asks for a page number or a bounding box, so you have citations with that. Looking through it, there's a pivoted matrix, a single long table, and a cross-page continuation. There are forms with checkboxes, a W-2, presentations, dense tables, electric bills, and a signature as well. This seems like a really practical dataset. I'm interested to see how the leaderboard turns out. I think it's brand new, so I wouldn't necessarily trust it yet, but hopefully in a week or two everyone will be testing their models against it and we can see where it shakes out.

articleReducto Blog09.01.26

Reducto launches r-1 document parsing model in preview

PRODUCT/API UPDATEMATERIAL

Reducto r-1 on watermarks, redlines, and visual diagrams

My take — Reducto just launched a new agentic OCR model called r-1. It's about one cent per page, and Reducto says it's about 20 percent better than their previous pipeline. I tried it out on a few documents that other models have had problems with to see how it would work. First, I tried it on a legal document that has this large watermark, which is critical for context and how to use this. Gemini with a generic transcription, Docling, and other approaches missed this, but r-1 did catch this and marked it as draft, not for execution. On a redlined document, it kept the strikethrough with the new change correct in context, which is great. But there were other models that did this well, like Chandra. On a rotated flowchart, a lot of models got the connections wrong. r-1 did as well, and Chandra and other approaches did too. I didn't find a silver bullet thing I could just pass this into without worrying about it. In general, I think r-1 is pretty strong. I ran it through a 22-question private eval, and it was really competitive. It actually got more right than giving the vanilla Pi harness Surya outputs or Chandra outputs. It's something worth considering if you are working with really high-fidelity forms or tables, redlines, or watermarks. But when it comes to visual diagrams, you still need to look carefully and make sure they're correct.

modelHugging Face09.01.26

inclusionAI releases ArmorOCR model family for grounded adversarial OCR

MODEL DROPNEW

articleDocling Blog09.01.26

Docling Integrates with Model Context Protocol and Introduces Reusable Agent Skills

PRODUCT/API UPDATEMATERIAL

X@UnstructuredIO08.28.26

Unstructured extracts code blocks as a distinct CodeSnippet element type

PRODUCT/API UPDATEMINOR

X@NielsRogge08.28.26

Papers with Code adds support for external, non-arXiv PDFs

PRODUCT/API UPDATEMATERIAL

Chandra gives Papers with Code full-text access

My take — Papers with Code is now using Chandra, thanks to Datalab, to accept non-arXiv PDFs for submissions. This is for their Chat with Research Paper feature. On Papers with Code, a page has an abstract, benchmark information, and, of course, links to code and Hugging Face as well. Understanding the paper takes quite a lot of time, and it's helpful to have a chatbot and an LLM to help with that. This is a good example of how, if you have a good document processing pipeline with a good OCR model, you can open it up to a lot more sources and work with a lot more data. I'm pretty excited about this one because it just means it's going to be a little bit easier and a little bit more accessible to learn what's going on with new papers and new releases.

articleLandingAI Blog08.13.26

LandingAI releases Snowflake streaming pipeline using ADE Gen2 and DPT-3

SOFTWARE RELEASENEW

LandingAI DPT-3: Extract document fields into SQLite

My take — LandingAI's ADE Gen2 releases DPT-3, a model that turns PDFs into text, page structure, and source locations. A separate extraction call returns whatever fields you ask for. I tried it on a loan estimate and a closing disclosure from my doc lab. I asked for estimated and final cash to close, plus estimated and final closing costs. My code loaded those fields into SQLite. Across both pages, all six dollar amounts matched the source. Cash to close went from $16,054 to $14,147.26, while closing costs went up from about $8,000 to about $9,700. The loan estimate's two final amounts weren't on the page and ended up empty in SQLite, just as we would expect. I think there's a lot of use cases for structured extraction and then putting it into a database like SQLite, but also Snowflake or Postgres or whatever you're using in production. A lot of these documents have tables or specific values that you want to be able to query or show in a dashboard all the time, and this seems like a great way to do it. If you try this out on your documents, let me know how it goes.

Document news

GitHub

Hugging Face

Web

X

YouTube

How Finch Legal Uses Reducto to Help Paralegals Manage 5x More Personal Injury Cases Reducto

New API Option Generates PDFs Passing Automated Accessibility Checks

LandingAI Introduces Split API to Separate Bundled Documents into Typed Sub-Documents

Chandra Open-Source Tool Accelerates STEM Braille Transcription Workflow

Davis County School District uses Chandra OCR to streamline STEM braille transcription

tencent/EVIE family released

Nutrient releases form-field-v1-state, an open-weight form field classifier

ParseBench v1.0.0 released with extensive document parsing pipelines and PyPI packaging

STAIR (STructure Aware Information Retriever): A novel dataset and LLM based retriever for document structure augmentation

LlamaIndex releases Turbo beta extraction tier for low-latency document processing

LlamaIndex and Kaggle launch ExtractBench document extraction leaderboard

Reducto launches r-1 document parsing model in preview

inclusionAI releases ArmorOCR model family for grounded adversarial OCR

Docling Integrates with Model Context Protocol and Introduces Reusable Agent Skills

Unstructured extracts code blocks as a distinct CodeSnippet element type

Papers with Code adds support for external, non-arXiv PDFs

LandingAI releases Snowflake streaming pipeline using ADE Gen2 and DPT-3