The country's first batch of Alibaba Damo Academy Document AI passed the authoritative evaluation of the Academy of Information and Communications Technology

On August 16, China Academy of Information and Communications Technology released the first round of evaluation results of intelligent document processing at the Trustworthy AI Summit. Alibaba Damo Academy’s document intelligence platform performed well and became the first document AI in the country to obtain "Trusted AI Certification" product. Document AI can identify and understand various complex documents such as bills, contracts, forms, etc. It is recognized as one of the highly difficult technologies in the industry. Alibaba DAMO Academy has built a complete technology stack and continues to lead the industry.

China Academy of Information and Communications Technology began to build a "trustworthy AI" evaluation system in 2018, and has gradually become an authoritative domestic artificial intelligence evaluation system, covering three categories: product service capabilities, application maturity and credible risks. This year, China Academy of Information and Communications Technology launched an evaluation for Intelligent Document Processing (IDP) for the first time. Through a comprehensive evaluation of more than 100 key indicators, Damo Academy Document AI received the highest level 5 evaluation in terms of technical capabilities, product capabilities, and application capabilities. , excellent performance. The Academy of Information and Communications Technology pointed out that Damo Academy’s document AI has complete functions, rich scenarios, wide industry coverage, high accuracy and generally high acceptability.

According to reports, Document AI is a further upgrade of OCR (optical character recognition) technology. Traditional OCR mainly targets text recognition in fixed formats and is difficult to deal with complex situations. Document AI can analyze various random layouts, identify hierarchical and structural relationships in documents, and even understand complex tables... Due to the complexity and variety of tasks, Document AI requires a deep integration of natural language processing and computer vision, and has always been recognized as one of the most difficult technologies in the industry. one.

Alibaba Damo Academy took the lead in proposing a multi-modal document information extraction solution based on graph models as early as 2019, leading the development direction of the industry; it has now built a complete document AI technology stack , in addition to the core document processing technology, it also includes underlying electronic document parsing, OCR and self-learning platforms. At the same time, DAMO Academy is also exploring the next generation of document intelligence technology, and has proposed the multi-modal document understanding model Bi-VLDoc. For the first time, it has achieved bidirectional vision-language through cross-supervision of different modal signals and forced mixed attention of different modalities. Accurate alignment, achieving the best model performance (SOTA) in four representative document understanding data sets in the industry.

DAMO Academy created a new SOTA on four representative data sets

It is understood that DAMO Academy Document AI It supports the automatic identification, extraction, classification, integration and verification of various document contents such as contracts, bills, and reports. It has been widely used in customs, legal, medical, financial and other industries, and is one of the important supporting technologies for enterprise digitalization. Take the customs declaration business as an example. It has been implemented in five major port areas including Shanghai and Ningbo, allowing corporate customs declaration personnel to avoid complicated manual entry. The relevant system has been in operation for more than two years and has processed more than two million customs declarations, improving efficiency for customs declaration companies by 3.5 times. In addition, DAMO Academy's document AI is also used in Braille recognition to translate Braille into Chinese characters and numbers. It has been implemented in schools for the blind in Zhejiang Province to help provide inclusive education.

AI automatically generates customs declaration form

AI translates Braille mathematics test papers

According to reports, Alibaba continues to focus on investing in document AI, including cutting-edge research on artificial intelligence. The latest "Cloud AI Developer Service Key Capabilities Report" by Gartner, an authoritative international research organization, shows that Alibaba ranks second in the world in the field of language AI, setting a record for the best results among Chinese companies.

The above is the detailed content of The country's first batch of Alibaba Damo Academy Document AI passed the authoritative evaluation of the Academy of Information and Communications Technology. For more information, please follow other related articles on the PHP Chinese website!