OCR / HTR - Text recognition
 

 

What is it?
 

Odoma is specialized in the high-quality recognition of text from printed and handwritten documents. These techniques are called Optical Character Recognition (OCR) and Handwritten Text Recognition (HTR) respectively. 

We have research and applied experience in OCR and HTR, using the latest deep neural network techniques.

Odoma is able to deal with non-standard sources including multiple layers of writing, degraded images, complex layouts, unconventional calligraphy, and more.

 

Results:
 

  • Rapidly search through massive amounts of text
     

  • Extract information from it (see Information extraction)
     

  • Develop new text-based services for staff and users
     

  • Extract complex, multi-layer textual content

Information extraction
 

 

​What is it?
 

Extract visual (e.g., pictures and drawings), structured (e.g., tables, references), and targeted (e.g., entities, events, numbers) from digital or digitized documents.

Thanks to our research experience, we can develop custom extraction algorithms dealing with complex multi-layer information, multiple data modalities (e.g., text and visual), as well as more standard scenarios.
 

Results:

  • Know what's inside your documents and make it searchable
     

  • Automatically populate knowledge bases with extracted information (see Knowledge bases below)
     

  • Develop new content-based services for staff and users
     

  • Create valuable datasets for research & development

Knowledge base
 

 

What is it?

Odoma support its clients in the design and development of rich knowledge bases, which can be populated with automatically extracted contents.

Such knowledge bases follow best standards and ontologies, can be developed using the latest database technologies (e.g., linked data and graph databases) as well as proven ones (e.g., relational databases).

Furthermore, Odoma is able to support you in the automation of the record linkage task, where entities and concepts are connected to authority records, either public or internal ones.


The development of structured and relational knowledge bases with information from large collections of data is an essential prerequisite for the development of powerful new applications, including search and navigation.
 

Results:

  • Turn unstructured data into knowledge bases: data becomes a usable asset
     

  • Develop knowledge bases following international standards for maximum interoperability
     

  • Automatically populate knowledge bases with AI techniques
     

  • Find every mention of entities or concepts of interest in your data
     

  • Develop powerful search and analytics applications relying on knowledge graphs

 

Explainable AI
 

What is it?

In complex and challenging scenarios, AI can be used to augment, rather than substitute, human expert decisions.

Odoma researches and develops techniques to support decision-making with systematic use of data analytics and explainable AI methods. 


These services are experimental and require a customized approach, tailored to a client's needs.
 

Results:
 

  • Augment expert decisions with data analytics and explanations of how AI models make decisions
     

  • Inform decisions in complex scenarios with systematic use of your data, while not automating them
     

  • Complex scenarios requiring expert decisions include: where data is available but partial or missing context; when legal constraints apply; when human expertise is not expressed in data

 

AI Workflows
 

What is it?
 

Data processing workflows are complex and full of dependencies, their partial or full automation with AI requires considering them in their entirety.

Data and automation flows must be designed, tested, and implemented in synergy.

Odoma is uniquely positioned to help its clients with creating customized AI workflows that can include any of the services we propose, and more.
 

Every workflow is different and requires a dedicated approach. We often suggest that our clients start with a preliminary assessment and pilot project, to prepare the ground for a greater effort. Please contact us to know more.

Results:

  • Re-think existing or new data and decision-intensive workflows in view of automating them with AI techniques
     

  • AI workflows have the potential to scale time-consuming tasks significantly, maintaining or improving their outcomes
     

  • AI workflows require a systematic design of data flows, which leads to the improvement of the quality and transparency of data procedures