Скачать или смотреть Document Analysis - Beyond OCR, d

Doc2answer enable a user to navigate through a dataset of complex documents. The problem we are addressing lies at the core of the digitisation process: normal document OCRs, when applied to complex datasets, are missing key contextual information.

The consequence is that the end-user is not able to fully capture the nuances of the data they are analysing.

To this end, we designed and create a context-aware, visual data analysis Software-as-a-Service (SaaS) platform.

Doc2Answer ingests visual content and links it together through a taxonomy of attributes. In particular, such taxonomy will relate attributes of three different kinds: 1) Personal Identification Information, 2) specific keywords, 3) layout and types of forms. In this way, we will enable a multi-attribute, context-aware search experience.

The core part of Doc2Answer is a set of AI algorithms able to automatically recognise visual attributes of interest.

It is worth noting that that: 1) context analysis is key: the most interesting attributes are handwritten corrections, stamps and structures of the forms, 2) bad quality of the paper can affect the performance of detection algorithms 3) document layout of cards can vary dramatically, but do bring added layers of information.