Korea Deep Learning Launches Mobile 'DEEP Agent' to Support On-Site Document Digitization

Korea Deep Learning announced on the 1st that it has launched the mobile app for its document AI agent, 'DEEP Agent,' and that it will help directly convert documents and images generated in the field into enterprise work data. 'DEEP Agent' is built on 'DEEP OCR' and 'DEEP Parser' and can extract text, field, and table information from documents and images, then structure the extracted data.

The mobile version stands out for extending existing web-based document processing functions to on-site operations. Field staff can take photos of documents, monitor screens, meters, business cards, and more with a smartphone, and the mobile app automatically extracts the necessary values and table structures from them.

The mobile app links images and processing results with the DEEP Agent web platform. As a result, headquarters staff no longer need to go through the process of sending photos via messenger or email and downloading them again, and they can check the results on the web without separately entering the required values.

Korea Deep Learning said that in existing field operations, work had become fragmented as data transfer, downloads, and re-entry were repeated after taking photos. The company said that with the mobile app, the process has been reduced from four steps — taking photos, transferring them, re-entering data, and checking results — to two steps: taking photos and checking results on the web.

The company said it had conducted advance deployments at 14 sites before launch, and that as a result, re-entry work decreased by 86%. It also said the app can be applied to a range of field tasks, including collecting documentation to verify pharmaceutical sales performance, reading gas meters, managing inspection checklists at manufacturing and facility sites, and registering business card information from sales visits.

Korea Deep Learning said that extraction criteria by task can be applied identically on the web and mobile versions. It explained that the operating model is one in which headquarters staff set the required items and table structures on the web, field staff collect data using the same template, and the results are checked against the same criteria.

It also said that when document formats or extraction criteria change, it uses DEEP Ops, its document AI operations platform. Through DEEP Ops, extraction criteria can be changed without separate rebuilding or development requests, performance before and after changes to the same document standard can be compared and verified, and changes can be reflected in web and mobile templates, the company said.

Korea Deep Learning is focusing on accumulating simple captured files from the field as data that can be used in enterprise AI and business systems. Based on this, it plans to expand the scope of DEEP Agent to field-centered tasks such as manufacturing, sales, and meter reading.

Kim Ji-hyun, CEO of Korea Deep Learning, said that if the work of a headquarters employee re-entering data is needed after a field employee takes a photo, then automation remains stuck at the capture stage. She added that the company will pursue a direction in which captured documents and extraction results are immediately linked on the web to reduce bottlenecks in the transfer and re-entry process and extend document data infrastructure to the field.

Source: TECHWORLD · Kim Seung-gi
Original: https://www.epnc.co.kr/news/articleView.html?idxno=407675

References

• https://www.epnc.co.kr/news/articleView.html?idxno=407675