vencortex® is excited to announce that our team received another grant funding for R&D of our decision augmentation technology by the Distr@l program as part of Hesse digital strategy for a project from 15.09.2020 to 14.09.2022.
The project DeepFeat (20_0032_2A) will apply deep learning to combine heterogeneous data for the use in enterprise contexts and was granted with a project volume of €625.218 (Grant €281.376). The project is led by our CTO Dr. Idan Shilon.
Data is a central resource for companies and is often referred to as the "new oil". Especially in the context of artificial intelligence (AI) and machine learning applications or machine learning applications, they are usually the decisive enabler for decision support in the company, automation or the development of new business models.
Currently, the greatest challenge for companies of all sizes is to create value from their own data. Since around 80% of all data in companies is unstructured (e.g., in texts, images, or customer feedback), companies are faced with the challenge of to make this data usable for machine learning processes. At the same time, this data is strongly context-dependent. For example, terms can be interpreted differently if they occur in different texts (e.g., different texts (e.g., customer feedback vs. contract texts). In addition current methods are usually optimized for only one type of data (e.g., text). In reality, different data types have to be combined with each other. This has the consequence that only 32% of all companies are actually able to use data in a way that creates value.
The goal of the project DeepFeat is the development of a framework and the prototypical implementation of a software application for pilot customers, which can combine categorical data (e.g., gender, colors etc.), numerical data (e.g., numbers) and unstructured data (e.g., images and text) by means of neural networks to create a latent feature space based on which machine learning models can be trained to solve highly complex problems in the enterprise context. The special feature of the solution developed in the DeepFeat project lies in the development of a method that is able to combine different data types in the enterprise context and to generate a latent feature space (features) with the help of Deep Learning methods, which is on the one hand contextualized and on the other hand can be integrated into a hierarchical enterprise ontology (knowledge graph).
While existing solutions are either limited to structured data or technical methods focus mainly on texts or images, the DeepFeat approach enables the combination of these data types with organizational data (e.g., system data) and categorical data, which are an elementary part of an enterprise ontology (knowledge graph).
Another key unique selling point of the DeepFeat solution is that it is delivered as a standardized software application, which makes it possible for companies of all sizes to easily combine to easily combine organizational data and thus leverage the value of data for the enterprise even without extensive data science knowledge. Up to now this has been Either only for corporations with large data science teams or enormously high-priced consulting projects possible. The DeepFeat solution thus enables easier access for SMEs to the use of (un)structured data in a business context.
More information can be found here.