VEGA is Affinda's Intelligent Document Processing Engine. VEGA understands and processes any document type, to automate previously human-led workflows.
- Can process any document type
- A range of powerful capabilities
- Leverages the latest advancements in AI
- ‘Human-in-the-loop’ functionality
- World-leading accuracy
- Processes in 50+ languages
Natural Language Processing
Our AI technology takes advantage of all the newest developments in artificial intelligence, machine learning, and natural language processing (NLP) to recognise and extract all the data from documents.
Affinda uses image recognition to successfully interpret the various sections of documents and parse them correctly. Then our NLP gets to work, using predictive models to understand the semantics behind the words and filter them into the right categories. The final product of this semantic search and data extraction is highly reliable structured data that can be easily integrated into any platform using our easy-to-use API.
The base models used to create our document-specific models have been trained on millions of documents of various types and importantly various languages. These models represent the latest and most advanced AI models available in NLP. That these models have been trained on documents in multiple languages means that our document-specific models will perform well across the wide set of languages out of the box. Our models perform well on 56 different languages, however further training on specific models can further improve performance.
While our document parsing model beats our competitors in competitive testing, we are constantly working on improving further to maintain our status as a leading-edge provider of document parsing services. The more data that our model receives, and the greater range of formats of each document type it sees, the better the model will perform. That means that Affinda's AI model can keep improving over time as we identify documents that aren't performing as well as expected, and the corrected data is fed back into the model as a feedback loop.