Discussion & outlook

Text Classification with Open Source Models

Five concluding thoughts

• We know it’s been a lot of theory 🤯, BUT understanding LLMs is helpful amid the AI revolution & for running the models, a basic understanding is enough.
• We know the code looks super complicated 😓, BUT loading and running the models via Transformers is pretty straightforward – as always, data wrangling is the mess, but Chatbots can help you here
• We know Python is scary 😱, BUT you can use it only to a minimal extent and quickly return to familiar territory
• We know commercial models are so tempting (Claude 🤤), BUT only Open-Source models guarantee data security and replicability (i.e., basic standards of our work)
• We know proper validation takes soooo much time 🫩, BUT just running a model is no ‘easy fix ’; it’s a black box, probably more dirty than quick and definitely not a scientific result you want your name attached to.

Thank you!
🥳

Please evaluate this workshop!

Contact

Felix Dietrich
Johannes Gutenberg Universität Mainz

felix.dietrich@uni-mainz.de
linkedin.com/in/felixdidi/

Daniel Possler
Hochschule für Musik, Theater und Medien Hannover

daniel.possler@ijk.hmtm-hannover.de
linkedin.com/in/dpossler/