Skip to content

Funded by

More information can be found at www.efre.gv.at

Project description

Digitalization of business processes results in the design and deployment of various information systems, such as search engines, databases, wikis, or file shares, to collect the knowledge and data needed to accomplish various tasks efficiently. However, these systems are usually storing data in formats that can only be interpreted by humans, e.g., images or texts. Therefore, users of these systems must process the retrieved information manually. Whenever multiple systems must be used to solve a problem, a user must manually establish a missing connection, i.e., query data from one system and transform it into the input format of another. Such operations can be automated using modern Artificial Intelligence (AI) methods, but their application in these scenarios is impossible since their training requires the existence of ground-truth examples. Unfortunately, the latter are not available due to the aforementioned reasons.

The ONTIS project aims to create tools that automate the creation of ground-truth data for the training of AI methods. Thus, in the knowledge-based part of the approach, the core notions of the application domain and their relations are formally described in an ontology, which should then be populated over the data stored in information systems. The goal is to define unambiguous mappings from any piece of data in an information system to a corresponding ontology concept, thus, providing the missing training examples. Manual inspection and annotation of ontology concepts in databases or file shares are problematic. Application of machine learning (ML) methods, such as natural language processing or image classification, can significantly simplify this task. However, the integration of ML with knowledge bases is currently widely researched by the AI community. Within ONTIS, we are aiming at the research and development of an annotation approach that can be applied in various domains, such as medicine or electrical engineering, by making the following contributions:

  1. enable formalization of a domain in form of an ontology;
  2. allow for annotation of images using ontology concepts and their subsequent storage, querying, and visualization within the ontology development framework; and
  3. provide assistance during the annotation process using pre-trained machine learning models.