JPMorgan Develops Generative Language Model DocLLM for Enterprise Document Analysis
Financial services giant JPMorgan announced the introduction of a new tool called DocLLM – a smart language model designed to understand various types of business documents.
These documents include forms, invoices, reports, and contracts, which often have complex information in both text and spatial layouts, according to the paper released.
It further mentioned that, unlike other similar models, DocLLM doesn’t rely on expensive image technology. Instead of using costly image-related technology, DocLLM focuses on understanding the structure of documents by identifying and defining rectangles (bounding boxes) around important text segments. These bounding boxes serve as a guide for the model to recognize and analyze the content within those specific areas.
The model has a unique feature called ‘disentangled spatial attention’ – which means the model can efficiently pay attention to and process information within these outlined areas separately, thereby helping it understand the relationship between text and layout in a document.
DocLLM is specifically good at handling documents with irregular layouts and different types of content. It learns to fill in missing text segments during its training, making it effective in dealing with diverse document structures.
DocLLM to Tacke Existing Challenges with Business Documents
In the realm of enterprise datasets, documents with intricate layouts such as invoices, receipts, contracts, orders, and forms play a substantial role. The automatic interpretation and analysis of these visually complex documents offer significant advantages, leading to the development of AI-driven solutions.
Despite the notable progress made by Document AI (DocAI) in tasks like extraction, classification, and question-answering, challenges persist in real-world applications. A performance gap exists, especially in terms of accuracy, reliability, contextual understanding, and the ability to generalize to unfamiliar domains.
To that end, JPMorgan comes up with DocLLM. According to the paper released, to train DocLLM, JPMorgan used data from two main sources: IIT-CDIP Test Collection 1.0 and DocBank. The first dataset includes over 5 million legal documents related to the tobacco industry in the 1990s, and the second has 500,000 documents with distinct layouts.
Tests show that DocLLM performs better than other similar models on various document-related tasks. It outshines equivalent models on 14 out of 16 datasets and proves its adaptability on 4 out of 5 new settings.
Looking forward, JPMorgan plans to improve DocLLM by incorporating vision-related features in a lightweight manner, aiming to enhance its capabilities even further.
Disclaimer
In line with the Trust Project guidelines, please note that the information provided on this page is not intended to be and should not be interpreted as legal, tax, investment, financial, or any other form of advice. It is important to only invest what you can afford to lose and to seek independent financial advice if you have any doubts. For further information, we suggest referring to the terms and conditions as well as the help and support pages provided by the issuer or advertiser. MetaversePost is committed to accurate, unbiased reporting, but market conditions are subject to change without notice.
About The Author
Kumar is an experienced Tech Journalist with a specialization in the dynamic intersections of AI/ML, marketing technology, and emerging fields such as crypto, blockchain, and NFTs. With over 3 years of experience in the industry, Kumar has established a proven track record in crafting compelling narratives, conducting insightful interviews, and delivering comprehensive insights. Kumar's expertise lies in producing high-impact content, including articles, reports, and research publications for prominent industry platforms. With a unique skill set that combines technical knowledge and storytelling, Kumar excels at communicating complex technological concepts to diverse audiences in a clear and engaging manner.
More articles
Kumar is an experienced Tech Journalist with a specialization in the dynamic intersections of AI/ML, marketing technology, and emerging fields such as crypto, blockchain, and NFTs. With over 3 years of experience in the industry, Kumar has established a proven track record in crafting compelling narratives, conducting insightful interviews, and delivering comprehensive insights. Kumar's expertise lies in producing high-impact content, including articles, reports, and research publications for prominent industry platforms. With a unique skill set that combines technical knowledge and storytelling, Kumar excels at communicating complex technological concepts to diverse audiences in a clear and engaging manner.
