=Paper= {{Paper |id=Vol-3617/paper-03 |storemode=property |title=Knowledge Discovery in the Age of LLMs |pdfUrl=https://ceur-ws.org/Vol-3617/paper-03.pdf |volume=Vol-3617 |authors=Jakub Zavrel |dblpUrl=https://dblp.org/rec/conf/birws/Zavrel23 }} ==Knowledge Discovery in the Age of LLMs== https://ceur-ws.org/Vol-3617/paper-03.pdf 
                                Knowledge Discovery in the Age of LLMs
                                Jakub Zavrel1
                                1
                                    Zeta Alpha, Amsterdam, The Netherlands


                                                                         Abstract
                                                                         Large Language Models (LLMs) like GPT-3 are quickly changing the way NLP is done, and hence also
                                                                         how NLP is done for the purpose of knowledge discovery in the academic literature. Tasks that have
                                                                         traditionally been done by specialized NLP models, like entity extraction, summarization, and question
                                                                         answering, can now all be prototyped, usually with high accuracy, using zero-shot or few-shot prompting
                                                                         of LLMs. For example, this allows us to extract highly accurate meta-data, such as up-to-date author
                                                                         and affiliation, for scientific impact analysis from recent scientific papers1 . The combination of LLMs
                                                                         and Information Retrieval systems has recently evolved into the paradigm of Retrieval Augmented
                                                                         Generation, which is one of the most promising approaches to use LLMs for Question Answering
                                                                         and to reduce hallucination, especially for data sets that were not accessible to LLMs during training,
                                                                         such as private data sources or very recent documents. Retrieval Augmented Generation can even be
                                                                         expanded to large document collections, like full conference proceedings, to quickly create conference
                                                                         summaries, blog-posts or tabular digests. At Zeta Alpha we are building a modern platform for scientific
                                                                         and enterprise knowledge discovery with neural search and generative language models at the core. In
                                                                         this talk, we show how we integrate LLMs with neural search to answer questions, explain documents
                                                                         while reading, summarize large document collections, and generate meta-data on the fly.




                                    1
                                      See: https://www.zeta-alpha.com/post/must-read-the-100-most-cited-ai-papers-in-2022
                                BIR 2023: 13th International Workshop on Bibliometric-enhanced Information Retrieval at ECIR 2023, April 2, 2023
                                $ zavrel@zeta-alpha.com (J. Zavrel)
                                                                       © 2023 Copyright for this paper by its authors.
                                                                       Use permitted under Creative Commons License Attribution 4.0 International (CC BY 4.0).
                                    CEUR
                                    Workshop
                                    Proceedings
                                                  http://ceur-ws.org
                                                  ISSN 1613-0073
                                                                       CEUR Workshop Proceedings (CEUR-WS.org)



                                                                                                                                         21




CEUR
                  ceur-ws.org
Workshop      ISSN 1613-0073
Proceedings