# PubMed-OCR: PMC Open Access OCR Annotations Source: https://www.youtube.com/watch?v=5_xZj7A8gCY Recap page: https://rapidrecap.app/video/5_xZj7A8gCY Generated: 2026-01-23T15:35:18.342+00:00 --- ## Quick Overview The Roots.AI team developed a new layout-aware OCR model that significantly outperforms Google's Cloud Vision OCR by accurately distinguishing between text and embedded visual elements like stamps and tables in scientific documents, achieving a 45% accuracy rate on complex data extraction tasks. **Key Points:** - The Roots.AI team created a new OCR model specifically designed to overcome the "plain text trap" that plagues standard OCR when processing scientific documents. - The new model achieves a 45% accuracy rate in extracting complex data, specifically successfully identifying and separating text from visual elements like stamps and tables. - Google's Cloud Vision OCR, which was used as a baseline, failed because it treated every page as a single image, leading to errors like extracting bounding boxes around non-textual elements. - The new model uses a layout-aware approach, grouping words based on their vertical position (like columns) to maintain contextual integrity. - A key failure case for standard OCR was accurately reading formulas and charts embedded within text, which the new model handles by visually processing the document structure. - The cost for processing 1.5 million pages of data using the new method was calculated to be $1.50 per 1,000 pages, significantly cheaper than anticipated costs for large-scale data processing. ![Screenshot at 00:22: The speaker introduces the core problem: asking if computers can read scientific documents, contrasting simple text extraction with the complexity of visual layout in scientific papers.](https://ss.rapidrecap.app/screens/5_xZj7A8gCY/00-00-22.jpg) **Context:** The AI Papers podcast discusses the development of a new Optical Character Recognition (OCR) technology by the team at Roots.AI, specifically addressing the challenge of accurately extracting structured data, like formulas and tables, from complex scientific documents such as those found on PubMed Central (PMC). The discussion contrasts this new layout-aware approach with the limitations of existing commercial tools like Google's Cloud Vision OCR. ## Detailed Analysis The discussion centers on the limitations of existing OCR technology, exemplified by Google Cloud Vision, when dealing with complex scientific documents. Standard OCR often falls into the "plain text trap," treating the document merely as a string of characters or a flat image, causing it to merge text across columns or pull in graphical elements like stamps and tables as text, resulting in nonsense. The team at Roots.AI developed a new layout-aware OCR model that addresses this by analyzing the document's visual structure. This new model achieved a 45% accuracy rate on complex data extraction, successfully identifying coordinates for words and separating foreground text from background elements like stamps (e.g., the confidential seal on an old document). This spatial awareness allows the model to correctly interpret layout structure, such as columns and tables, which generic OCR tools fail to do, leading to far more reliable data extraction, particularly for highly technical content like medical articles. ### Introduction to OCR Challenges - Tackling the hardest problems in artificial intelligence: reading scientific documents - The issue with plain text extraction - The problem is extracting data from visual documents, not just text messages. ### The Problem with Current Tools - Standard OCR fails by treating documents as flat text strings - Google's Cloud Vision OCR incorrectly groups text across columns - This leads to garbled output when reading formulas or charts. ### Roots.AI Solution - Developing a layout-aware OCR model - The model uses geometric rules (like vertical position) to group words correctly - It extracts the hidden digital text layer while respecting visual context. ### Performance Metrics - The new model achieved 45% accuracy on complex extractions - It successfully separates text from visual elements like stamps and tables - It correctly interprets formulas and tables, which older models miss. ### Future Implications - The goal is to create the next generation of layout-aware OCR technology - This improved accuracy builds trust and allows for better data processing in specialized fields like healthcare. ![Screenshot at 00:13: The speaker begins explaining the topic: ChatGPT writing a poem versus the complexities of reading scientific documents.](https://ss.rapidrecap.app/screens/5_xZj7A8gCY/00-00-13.jpg) ![Screenshot at 00:33: The speaker illustrates the problem by noting that if you strip away layout information, the document becomes meaningless to a standard parser.](https://ss.rapidrecap.app/screens/5_xZj7A8gCY/00-00-33.jpg) ![Screenshot at 01:44: The speaker explains that meaning is derived from layout, using the example of a number in a table correlating to a column header.](https://ss.rapidrecap.app/screens/5_xZj7A8gCY/00-01-44.jpg) ![Screenshot at 02:28: The speaker summarizes the failure of standard parsers: 'Garbage in, garbage out.'](https://ss.rapidrecap.app/screens/5_xZj7A8gCY/00-02-28.jpg) ![Screenshot at 06:55: The speaker introduces the classic OCR challenge: avoiding the 'clutter' caused by visual elements like stamps overlapping text.](https://ss.rapidrecap.app/screens/5_xZj7A8gCY/00-06-55.jpg)