Semantic technologies help to establish common grounds for IoT data and give means for doing analysis and developing, for instance, machine learning approaches on top. Real added value comes when the data is not just data, but has meaning and correlations and context.
SEMANTiCS: You have been recently working on the topic of Industry 4.0 and IoT. Where do you see the place for semantic technologies in the area of cyber-physical systems?
Maria Maleshkova: With the constantly increasing number of sensors and mobile devices being used, the employment of semantics becomes an absolute necessity. IoT data brings a lot of potential but it comes with a number of challenges that need to be addressed. We see heterogeneous data, with different frequencies, made available in different ways, in greater volumes, of which only small portions have a certain significance. Therefore providing a basis for integration and being able to make sense of the data is absolutely crucial, and this is precisely where semantics comes into place. Not only do semantic technologies help to establish common grounds for the data but they also give the means for doing analysis and developing, for instance, machine learning approaches on top. The current downside of the solutions is that they are custom for each industrial domain and for each individual use case. Here we need to do more in terms of enabling reuse and wider adoption.
SEMANTiCS: One of the hot topics in computer science is the increasing convergence of semantic computing, machine / deep learning and language processing towards the next generation of AI systems. What is your opinion on this? Just another AI spring or is AI finally here to stay?
Maria Maleshkova: I think that this is the natural next step in the evolution of AI-solutions. In order to provide a good AI system, you foremost need data and the means to structure, understand and make sense of this data before actually applying the AI-approach. This is why language processing and semantic technologies fit so well with AI-solutions. Working with the raw data blindly is alway an option but the real added value comes when the data is not just data, but has meaning and correlations and context. I am convinced that we will be seeing more and more of this fusion of technologies and approaches because the synergies are extremely beneficial. AI is here to stay! We see this clearly both in research and in the industrial domains. However, we will very soon have a problem with educating a sufficient number of AI-experts. Currently, we are seeing an extreme demand for data scientists and the next natural level would be the need for AI-experts. We are in no way prepared to address this need and this is a very big problem. The new technologies can be adopted in the markets only when you have the experts to drive this adoption, and AI experts will very soon become a scarce resource.
SEMANTiCS: Together with Ilaria Tiddi from the University of Amsterdam, you are this year’s Poster and Demo Chair. Any recommendations for submitters that make a great poster?
A great poster catches the attention of the viewer immediately. We react to the visual elements first, so a prominent picture, a distinctively written title, a logo or a bold statement are some options to appeal to the audience right away. A great poster also transports the main messages of the work clearly, with a few images and bullet points, and finally it also invites discussion. It follows a clear structure for positioning the content on the page and, usually, less is more. So leaving some white space is not a bad thing. I like to think of posters as supporting visuals, which enable the researchers to present their work. So at the end of the day, the researcher is the one delivering the content and the poster is there only as an assisting tool.
