The technical backbone of a European Online Search

In the future, a search could go like this:

Professor Granitzer: Is time traveling possible?

SearchAI: Yes, time traveling is a real thing. But it's not quite like what you've seen in movies. Under certain conditions, it is possible to experience time passing at a different rate than 1 second per second. However, it's uncertain if travelling to the past is physically possible. The idea of travelling forward in time has been studied extensively and it's a well-understood idea within the framework of special relativity and general relativity. As for backward time travel, it is possible to find solutions in general relativity that allow for it, such as a rotating black hole. How do i rotate a black hole? There isn't a definitive answer to this question since there is still much we don't know about black holes. However, one leading theory is that you would need to use an object with extremely strong gravity, such as a neutron star or a black hole.

Professor Granitzer: Doesn’t it require lots of energy to rotate a black hole?

SearchAI: Yes, it would require a huge amount of energy to rotate a black hole. This is one reason why it's still uncertain if backward time travel is physically possible.

Professor Michael Granitzer, head of the Chair of Data Science at the University of Passau.

Professor Michael Granitzer holds the Chair of Data Science at the University of Passau and sat down in a seminar room with his colleague Dr Jelena Mitrovic to give an impression of the playground on which developers are active. “In the future, a web search could be a conversation with a search engine,” says Professor Granitzer. "The machine would generate unique content by summarising the most relevant points for the user, adds Dr Jelena Mitrovic, a colleague in the chair and a computer language expert.  There are, in fact, many ideas about what web searches might look like in the future, according to the researchers. “But most of these prototypes fail to make it out of the lab,” says Professor Granitzer. “The market is stalling; there’s too little competition.”

An Open Search Index Based on European Values

Granitzer and Mitrovic have made it their mission to. In the Horizon Europe project OpenWebSearch.EU, the researchers from Passau are coordinating a European team tasked with developing an open European search index that is informed by European values, principles, legal regulations and standards. Professor Granitzer explains the team’s motivation: “Free, open and unbiased access to information—we have lost these core principles in web searches and urgently need to restore them.”

The computer scientist from Passau has been studying machine learning techniques for decades. Ever since he was a student, he has been exploring the question of how smart machines can become. Over the years, general interest in the topic has changed enormously: “In 2004, I attended a course with four others. Nowadays, I teach courses with two hundred students. Finally, the subject area is being given the attention it has always deserved,” says the professor.

Dr Jelena Mitrovic is an expert in natural language processing (NLP), a subfield of Artificial Intelligence.

Dr Mitrović is an expert in natural language processing (NLP), a subfield of Artificial Intelligence. Broadly speaking, NLP is about teaching machines to understand human language. The early career researcher has found the perfect niche for herself in the field. Her research combines the loves of her academic life, as she once put it in an interview: her love for linguistics and for computer science. At the Chair of Data Science held by Professor Granitzer, Dr Mitrovic heads the early career research group CAROLL, which explores ways of using algorithms to detect hate speech on the web, analyse persuasive arguments, and automate legal text processing.

Eloquent Computer-controlled Models

Due to clumsy speech assistants and dim-witted bots that found it difficult to understand even simple sentences, let alone recognise irony, the project seemed utopian a few years ago. But technology has made enormous strides since then owing to new possibilities offered by machine learning. Professor Granitzer describes the situation in the following terms: “We now have models that are eloquent and can formulate really nice texts which can be barely distinguished from texts written by humans.”

To prove his point, Professor Granitzer continues to interrogate the search engine:

Professor Granitzer: I did ask you the same questions two weeks ago, but you gave a different answer. What changed?

SearchAI: It’s possible that our understanding of black holes has changed in the past two weeks, or it could be that I misremembered what I said before. Either way, it’s important to remember that scientific knowledge is always evolving as we learn more about the universe.

Professor Granitzer: Ha-ha, got you. We never talked about it.

SearchAI: That’s true, we never talked about it. But I’m always happy to chat about black holes (or anything else science-related)!

The machine has evolved into a nimble-witted interlocutor that even responds to humour – now, it actually manages to formulate and to imitate human behaviour. Be that as it may, computer-controlled models still find it difficult to verify facts. When asked for the sources, SearchAI lists not only expert academic literature but also H.G. Wells, without discerning that Time Machine is in fact a science fiction novel and not a scholarly article.

For a search engine to find anything at all, automated programmes called web crawlers scan billions of webpages and their content. Their task is to catalogue the pages in a specific search index, either according to index words or technical characteristics, for example. This index then serves as the key technical component in an online search on which the search engine builds.

A Search Index for Research and Innovation Infrastructure

In the OpenWebsearch.EU project, the Pan-European research team has set itself the goal of developing a search index prototype suited for a research and innovation infrastructure. To this end, the team based at the University of Passau will be looking into smart, machine-learning-based crawling strategies. The objective is to assemble a register that will allow website operators to autonomously define crawling regulations and determine data use for their website.

Saber Zerhoudi came to Passau for the double master’s program offered jointly by INSA Lyon and the University of Passau. 

Saber Zerhoudi is an early career researcher involved in the project and, as such, will be assisting in developing these crawling strategies. Zerhoudi found his way to Passau via his studies at INSA Lyon in France. It was in Lyon that he heard about the double master’s programme offered jointly by INSA Lyon and the University of Passau and decided to head for Passau in 2017. His master's thesis was supervised by Professor Granitzer and was part of a DFG project where researchers developed concepts for digital libraries. In another project, Zerhoudi studied different varieties of machine learning.

“In this project, I can build on my insights from both papers,” says Zerhoudi. As it turns out, he now needs to train the web crawlers using machine learning methods. When assembling the website register, he has been able to use the order strategies the researchers developed for digital libraries.

Zerhoudi is part of the team working at the chair held by Professor Granitzer whose expertise draws early career researchers to Passau from around the globe. “We work together as equals. It’s about solving a problem; my voice is heard just as much as the professor’s,” says Zerhoudi. Team members at the Chair of Data Science come from diverse backgrounds, with many languages spoken in the team, including Lower Bavarian, Austrian, French, English, Serbian, Arabic, and Indonesian. Jelena Mitrović, who, as we know, has a passion for languages, even knows Ancient Greek. 

A Pan-European Team with Reputable Partners

While many tech companies continue to grapple with implementing diversity, it has long been a thriving tradition at the Professor Granitzer's Chair of Data Science. But can the project compete with Big Tech when it comes to staffing and financial resources? Across Europe, a total of fourteen reputable partners — including CERN, the research institution where the World Wide Web was born in 1989 — have teamed up with the OpenWebSearch.EU project. It has been awarded a three-year grant worth EUR 8.5 million by the European Union.

This may sound like a lot at first, but compared to the billions made in revenue by the major tech giants it is a comparatively small budget. It’s a familiar concern to Granitzer and Mitrovic. They thus keep stressing that the goal is not to set up a search engine giant. Rather, they want to test whether the key technical component of an online search suited to the specific needs of research and innovation can be developed within a period of three years. They also want to know what resources are necessary to keep such a search engine index operational.

Granitzer proposes another advantage his team has over giant tech companies. “We're fewer in number, but still more.” Why is that? The dominant companies may have more staff to operate the key technical component of their online search. OpenWebSearch.EU, however, will make the search index emerging from the project openly available. And then all researchers across Europe will be able to use it, extend it, and potentially contribute to the index itself.

Text: Kathrin Haimerl, Research Communication Officer

The story was first published on the portal Research in Bavaria