Recently, the semi-annual corporate bond report of Guangdong Guangxin Holding Group Co., Ltd. was released, which contained shocking content: On page 44 of the report, in the information section on important subsidiaries or joint-stock companies, the company name was written as "Then what else?" Let’s stay at my brother’s house.” This incident subsequently triggered heated discussions in the market. It is rumored that the person who wrote the semi-annual report of Guangxin Holdings’ bonds was a female intern at a financial company.
Public information shows that Guangdong Guangxin Holding Group Co., Ltd. is a wholly-owned enterprise of the Guangdong Provincial Government. Its predecessor was Guangdong Foreign Trade Group Co., Ltd. established in 2000. In 2011, it was renamed Guangdong Guangxin Holding Group Co., Ltd. Guangxin Holding Group is backed by Guangdong's state-owned assets. On August 2 this year, the Fortune Global 500 list was released. Guangdong Guangxin Holding Group Co., Ltd. appeared on the list for the first time, becoming the first provincial state-owned enterprise in Guangdong to appear on the list. Ranked No. 427 on the Fortune list).
When a Fortune 500 company issued a report and listed its subsidiary incorrectly, it was obviously a human error. The editor has collected a series of similar situations, which can be described as "heart-wrenching"...
According to a reporter from the Financial Associated Press, since August, A-share listed companies have made 48 correction announcements to their 2023 semi-annual reports. Among them, there are accounting errors in financial data such as assets and profits, and there are earnings per share in the semi-annual reports. The data is incorrect, and some content is entered incorrectly due to staff negligence, etc...
Source: Financial Associated Press
With the continuous advancement of digital transformation, the amount of enterprise data has shown explosive growth.
IDC Global DataSphere shows that by 2026, the total amount of global structured and unstructured data is expected to reach 221.2ZB. In the rapidly growing data, unstructured data is difficult and time-consuming to process due to complex formats and diverse standards. Facing massive unstructured data, manual processing will face huge challenges without automated solutions.
The purpose of document review is to reduce the risk of error management and control. If it relies heavily on manual labor, it will affect the organization's operational efficiency.
Especially now that fast pace, high efficiency, and fragmentation are common people's living and working conditions, how can we help document workers who are originally immersed in "reading" hundreds of thousands of words every day to quickly decompress these long documents?
Even, not just documents, but different types of documents such as contracts, bills, forms, reports, legal documents, etc., can it quickly, accurately and clearly help business personnel obtain the key information they want?
In response to these needs, IDP (Intelligent Document Process, hereinafter referred to as IDP) came into being. IDP can extract key information from semi-structured/unstructured data to form structured data, further realizing end-to-end automation of business processes and effectively improving the efficiency and accuracy of document management.
To reduce errors and improve efficiency, why not use IDP?
1. Intelligent document digital employees implement IDP to enable human-machine collaboration in document review
Faced with the above-mentioned business pain points and capability requirements, Shishi Intelligent uses its domestically-developed AI capabilities and powerful pre-training models to create an artificial intelligence solution in document analysis and review scenarios - Shishi Intelligent IDP, which enables content-intensive and long-length solutions . , the analysis and review of unstructured documents can be widely used in various contract, document, article and other review scenarios in various industries such as finance, manufacturing, communications, law, auditing, media, government, etc. , to achieve full-process empowered document management.
As an intelligent product that integrates various algorithms and business scenarios, Real Intelligent IDP has accumulated many AI capabilities, such as OCR license recognition, form recognition, keyword extraction, document comparison, document review, cross-document cross-review, etc. , combined with RPA automated operations, greatly improves the efficiency and accuracy of enterprise document processing. No matter which industry you are in or what business scenario you face, intelligent IDP can bring an efficient, stable, and smooth business experience:
Typical Scenario 1: Table Recognition and Analysis
In fact, IDP automatically identifies and extracts paper information and synchronizes it to the system, which can realize automatic information entry, external credit inquiry and information integration. For example, form information extraction, ID document information extraction, system information automatic capture, etc. can extract table titles and contents from various types of files. (doc .docx .wps .pdf .txt .jpg .jpeg .png .tiff .tif, etc.)
Typical Scenario 2: Bank Credit Review
In fact, IDP extracts important information from various documents, (such as financial report analysis, compliance review such as contract subject names, image data and system information verification, etc.) It also supports multi-format text such as word, pdf, pictures, scanned documents, etc. The accuracy rate of keyword information extraction is as high as over 95%. Assist bank approval personnel to determine the authenticity, consistency, compliance, etc. of loan information, and automatically complete information review and verification.
Typical scenario three: daily contract review
Really IDP has a variety of commonly used contract reviews and common risk points built-in, combined with custom risk points and logical operations, covering more than 95% of contract risks (realizing difference points comparison between text in multiple formats such as word, pdf, pictures, scanned documents, etc. , including punctuation, headers and footers, etc., the comparison accuracy is as high as 99.9% or more.), and can automatically generate comparison results reports to help companies quickly discover potential risks and give modification suggestions (supports review clause display, original text risk point positioning) , customized risk points, logical operation verification, etc.), strengthen risk control management, and protect the development of enterprises.
Typical Scenario 4: Analysis of Corporate Financial Reports
Supports the uploading of financial reports in various formats such as documents and PDFs. Through structured analysis and integration with financial linkage verification, the risk review, result modification and output of financial data are completed to ensure the health and quality of financial reports.
2. Advanced real IDP4.0: Chat-IDP talks to the document and lets the document speak for itself
As big language models usher in a new wave of technology, the application ecology of intelligent document review will become more complete, and model-driven development paradigm changes are rapidly taking shape.
On this basis, Real Intelligence further combines the capabilities of self-developed large models and continues to upgrade. At the seventh product launch conference, built by Real Intelligence’s leading AI capabilities and relying on core technologies such as Optical Character Recognition (OCR), Natural Language Processing (NLP), and Large Language Model (LLM), Real Intelligence created IDP4 The most representative experience in version .0 - Chat-IDP.
This is the first intelligent document digital employee that can talk to documents, achieving panoramic analysis of documents. Because of this foundation, document workers now have a work partner who can chat with AI: through Chat-IDP By talking directly to the document, you can continue to automatically analyze dense, long, and unstructured documents.
Generally speaking, in document work, there will be documents in multiple formats (pictures, PDF, word, etc.). The documents will have text descriptions, illustrations, tables, so based on multiple models, comprehensive analysis and understanding of complex documents can be achieved. is compulsory. This also provides innate conditions and foundation for conversational document processing.
Now, after users log in to Chat-IDP, they only need to upload files, and Chat-IDP can automatically understand various contents such as contracts, documents, books, etc., and is compatible with multiple formats such as Word, PDF, and scanned documents. Users can directly ask questions and communicate by selecting content on the original text, realizing functions such as key information search, rewriting and rewriting, logical operations, error correction and translation, etc., reducing the sense of fragmentation caused by jumping back and forth between documents and other AI function platforms.
The current ChatIDP product also integrates RPA process automation capabilities at the bottom layer (to realize automatic acquisition of different types of documents from multiple channels, and also to realize automatic backfilling after document processing), and integrates scene customization capabilities (to realize separate extraction of different business scenarios) , and cross-compliance audits). Combined with the TARS vertical large model developed by Real Intelligence, at the business level, it covers the entire process automation of document work, from the entire process of document collection, reading, understanding, review, and processing, to end-to-end automation coverage.
Redefining the document review model, I believe that Chat-IDP will not only be known and used by more people as an innovative intelligent document office product, but will also help the intelligent development of technology in the field of document review and contribute to the high-quality development of thousands of industries. Provide a pair of "fiery eyes".