Mojibake! What the h—ck happened to these strings? by Robyn Speer Our natural-language systems need to be prepared to deal with real-world text. But sometimes real-world text says something like “Merci de t‚Äö√†√∂¬¨¬©l‚Äö√†√∂¬¨¬©charger le plug-in”, and that’s terrible. I made a Python module called “ftfy” that can solve the puzzle of what happened to Unicode text like this, and in many cases undo it. I want to talk briefly (without too much blame) about why mojibake happens, how we can fix it and maybe prevent it, and why this isn’t a job for machine learning.pens, how we can fix it and maybe prevent it, and why this isn’t a job for machine learning. Robyn Speer is the developer of the multilingual knowledge graph ConceptNet, as well as several open source projects including ftfy. A decade ago, she co-founded an NLP startup called Luminoso. Now, in her increased free time, she plays and modifies randomizers for retro video games.