On May 30 I gave a plenary talk at the ICAME conference in Koblenz, where I discussed the changing meanings of AI and its role in corpus linguistics.
While the recent hype around generative LLMs has brought AI into the spotlight, corpus linguistics has relied on AI methods for decades, from part-of-speech tagging to word embeddings and other NLP tools. I showed how the meaning of “AI” itself has shifted, increasingly becoming associated with generative LLMs.
The second part of my talk focused on a more practical question: can we use LLMs responsibly in corpus linguistic research, given all the problems with transparency, reproducibility, copyright, privacy, fairness and environmental impact? I presented a case study of a very complex annotation task showing that at least some of these concerns can be addressed if we use open-weight, locally deployable models.
URL of the conference: https://wp.uni-koblenz.de/icame47/