New professor Suzan Verberne aims to bring large language models and search engines closer together
Suzan Verberne has been appointed professor of Natural Language Processing at the Leiden Institute of Advanced Computer Science (LIACS) from 1 October. Verberne has been at LIACS since 2017 as group leader of the Text Mining and Retrieval group.
Natural Language Processing (NLP) is the field of artificial intelligence that aims to enable computers to understand, interpret and generate human language in a way that is meaningful and contextually relevant. NLP has taken off in recent years with the development of large language models, the best-known example of which is chatGPT. These language models can generate fluent texts and bring all kinds of opportunities and challenges in research and practice.
The future of AI
Verberne's research mainly focuses on finding information in large text collections. She and her group are working on methods to improve search engines, as well as more specialised text mining methods. Think of discovering side effects mentioned by cancer patients on discussion forums, or finding the right answers to legal questions. Her expertise covers the interface of NLP and machine learning, areas that are crucial for shaping future AI technologies.
Practical applications
Verberne played an important role in several pioneering projects throughout her career. She developed innovative algorithms for unlocking information in many different domains, from archaeology to healthcare and from legal texts to posts on Facebook. She actively collaborates with researchers in other fields, for example in projects on recognising diverse opinions in news reports, analysing coverage of cancer screening, and identifying relationships between patents and scientific research.
Future challenges
"Large language models like chatGPT are now used by almost everyone for many different purposes: as writing aids, programming aids, creative assistants, and to search and interpret information," Verberne says. "It is therefore important to conduct research on these models: how can we ensure that the text they produce is factually correct and free of free of bias and toxicity? How can we build reliable and transparent chatbots with large language models, and how can we adapt them to specific areas without using large amounts of computing power? These are important questions in my group's current research.'